TMDB Movie Data Exploration
Introduction
The goal of the project is to derive insights on the TMDB movie datset and perform regression models to predict revenue of the movie. This model could be leveraged by production companies for making go/no-go screening decisions.
TMDB Movie Dataset available on Kaggle. Link: https://www.kaggle.com/tmdb/tmdb-movie-metadata
Load Packages
library(plyr) #data manipulation
library(tidyverse)# data manipulation
library(formattable)# table
library(splitstackshape) # split columns
library(jsonlite) #JSON format
library(wordcloud) #wordcloud
library(RColorBrewer) # Color Theme
library(ggthemes) #Themes for plot
library(tm) # Sentiment Analysis
library(RSentiment) # Sentiment Analysis
library(zoo) # Time
library(stringr) #String Manipulation
library(ggplot2)
library(VIM)
library(mice)
library(vcd)
require(car)
library(tabplot)
library(PerformanceAnalytics)
library(MASS)
library(glmnet)
library(dplyr)Load TMDB datset
movie=read_csv("tmdb_5000_movies.csv",col_names=TRUE,na="NA")
credit=read_csv("tmdb_5000_credits.csv",col_names=TRUE,na="NA")The Movie and credit data contain columns of nested JSON which need to be split into separate columns for accurate analysis. In the movie dataset, popularity means the number of views of the movie in the website and vote_average tells us about the movie rating.
glimpse(movie)Observations: 4,803
Variables: 20
$ budget <int> 237000000, 300000000, 245000000, 250000000, 260000000, 258000000, 260000000, 280000000, 250000000, 25...
$ genres <chr> "[{\"id\": 28, \"name\": \"Action\"}, {\"id\": 12, \"name\": \"Adventure\"}, {\"id\": 14, \"name\": \...
$ homepage <chr> "http://www.avatarmovie.com/", "http://disney.go.com/disneypictures/pirates/", "http://www.sonypictur...
$ id <int> 19995, 285, 206647, 49026, 49529, 559, 38757, 99861, 767, 209112, 1452, 10764, 58, 57201, 49521, 2454...
$ keywords <chr> "[{\"id\": 1463, \"name\": \"culture clash\"}, {\"id\": 2964, \"name\": \"future\"}, {\"id\": 3386, \...
$ original_language <chr> "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en",...
$ original_title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carte...
$ overview <chr> "In the 22nd century, a paraplegic Marine is dispatched to the moon Pandora on a unique mission, but ...
$ popularity <dbl> 150.437577, 139.082615, 107.376788, 112.312950, 43.926995, 115.699814, 48.681969, 134.279229, 98.8856...
$ production_companies <chr> "[{\"name\": \"Ingenious Film Partners\", \"id\": 289}, {\"name\": \"Twentieth Century Fox Film Corpo...
$ production_countries <chr> "[{\"iso_3166_1\": \"US\", \"name\": \"United States of America\"}, {\"iso_3166_1\": \"GB\", \"name\"...
$ release_date <date> 2009-12-10, 2007-05-19, 2015-10-26, 2012-07-16, 2012-03-07, 2007-05-01, 2010-11-24, 2015-04-22, 2009...
$ revenue <dbl> 2787965087, 961000000, 880674609, 1084939099, 284139100, 890871626, 591794936, 1405403694, 933959197,...
$ runtime <int> 162, 169, 148, 165, 132, 139, 100, 141, 153, 151, 154, 106, 151, 149, 143, 150, 143, 136, 106, 144, 1...
$ spoken_languages <chr> "[{\"iso_639_1\": \"en\", \"name\": \"English\"}, {\"iso_639_1\": \"es\", \"name\": \"Espa\\u00f1ol\"...
$ status <chr> "Released", "Released", "Released", "Released", "Released", "Released", "Released", "Released", "Rele...
$ tagline <chr> "Enter the World of Pandora.", "At the end of the world, the adventure begins.", "A Plan No One Escap...
$ title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carte...
$ vote_average <dbl> 7.2, 6.9, 6.3, 7.6, 6.1, 5.9, 7.4, 7.3, 7.4, 5.7, 5.4, 6.1, 7.0, 5.9, 6.5, 6.3, 7.4, 6.4, 6.2, 7.1, 6...
$ vote_count <int> 11800, 4500, 4466, 9106, 2124, 3576, 3330, 6767, 5293, 7004, 1400, 2965, 5246, 2311, 6359, 1630, 1177...glimpse(credit)Observations: 4,803
Variables: 4
$ movie_id <int> 19995, 285, 206647, 49026, 49529, 559, 38757, 99861, 767, 209112, 1452, 10764, 58, 57201, 49521, 2454, 24428, 186...
$ title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carter", "Spider-...
$ cast <chr> "[{\"cast_id\": 242, \"character\": \"Jake Sully\", \"credit_id\": \"5602a8a7c3a3685532001c9a\", \"gender\": 2, \...
$ crew <chr> "[{\"credit_id\": \"52fe48009251416c750aca23\", \"department\": \"Editing\", \"gender\": 0, \"id\": 1721, \"job\"...1. Data Exploratory Analysis
genredf=movie %>% filter(nchar(genres)>2) %>% mutate(js=lapply(genres,fromJSON)) %>% unnest(js) %>% dplyr::select(id,title,genre=name) #Convert JSON format into data frame
slice(genredf)1.1. Wordcloud: Genre Representation
#A look at the genre variety in our dataset
df <- as.data.frame(table(genredf$genre))
df2 <- with(df,df[order(Freq,decreasing = TRUE),])
#wordcloud
wordcloud(words=df2$Var1,freq=df2$Freq,min.freq=100,max.words = 20,random.order=FALSE,random.color=TRUE,rot.per=0.35,colors = brewer.pal(20,"Dark2"),scale=c(5,.2))n too large, allowed maximum for palette Dark2 is 8
Returning the palette you asked for with that many colors
1.2. Wordcloud: Production Companies
production=movie %>% filter(nchar(production_companies)>2) %>% mutate(js=lapply(production_companies,fromJSON)) %>% unnest(js) %>% dplyr::select(budget,revenue,company=name)
lapply(production,class)$budget
[1] "integer"
$revenue
[1] "numeric"
$company
[1] "character"df <- as.data.frame(table(production$company))
df2 <- with(df,df[order(Freq,decreasing = TRUE),])
df2wordcloud(words=df2$Var1,freq=df2$Freq,min.freq=50,max.words = 25,color =rainbow(7),scale=c(3,0.5))
1.3. Which year has seen maximum release of movies ?
movie$Year=as.factor(format(movie$release_date,"%Y"))
movie$Date=as.factor(format(movie$release_date,"%d"))
movie$month=month.abb[(as.factor(format(movie$release_date,"%m")))]
df <- as.data.frame(table(movie$month))
df2 <- with(df,df[order(Freq,decreasing = TRUE),])
df2df2 %>%
ggplot(aes(reorder(Var1,Freq),Freq,fill=Var1))+geom_bar(stat="identity")+theme(plot.title=element_text(size=14,face="italic",colour="red"),axis.text.x = element_text(angle=90),legend.position="none")+labs(x="",y="Total number of movies released",title="Number of Movies Releases per month")+coord_flip()+geom_label(aes(label=Freq))
1.4. Does high budget movie necessarily mean high popularity among viewers?
db=movie %>% left_join(credit,by=c("id"="movie_id"))
db_credit=db %>% filter(nchar(cast)>2) %>% mutate(js=lapply(cast,fromJSON)) %>% unnest(js)get_cor <- function(df){
m <- cor(df$x,df$y, use="pairwise.complete.obs");
eq <- substitute(expr=r==cor,env=list(cor=format(m, digits = 4)))
return(as.character(as.expression(eq) ))
}
temp=db_credit %>% dplyr::select(budget,popularity) %>% distinct()
ggplot(temp,aes(budget,popularity))+stat_bin_hex(bins=15)+scale_fill_distiller(palette="Spectral")+stat_smooth(method="lm",color="orchid",size=2)+scale_x_continuous(labels=scales::comma)
temp=db_credit %>% dplyr::select(budget,vote_count) %>% distinct()
ggplot(temp,aes(budget,vote_count))+stat_bin_hex(bins=15)+scale_fill_distiller(palette="Spectral")+stat_smooth(method="lm",color="orchid",size=2)+scale_x_continuous(labels=scales::comma)
temp=db_credit %>% dplyr::select(budget,revenue) %>% distinct()
ggplot(temp,aes(budget,revenue))+stat_bin_hex(bins=15)+scale_fill_distiller(palette="Spectral")+stat_smooth(method="lm",color="orchid",size=2)+scale_x_continuous(labels=scales::comma)
From the above plots we could realise that even when production companies spend lot of money on the movie it will not yield the revenue unless important features like story and direction make an impact to the audience.
1.5. Comparing Revenue with vote_count, popularity and runtime
ggplot(db_credit, aes(x =runtime, y =revenue))+
geom_point(size=1, aes(colour=vote_count)) +
labs(title = "Runtime Vs. Revenue and vote count",
x = "Runtime", y = "Revenue")
ggplot(db_credit, aes(x =popularity, y =revenue))+
geom_point(size=1, aes(colour=vote_count)) +
labs(title = "Popularity Vs. Revenue and vote count",
x = "popularity", y = "Revenue")
ggplot(aes(x = log(revenue), y = log(budget)), data = db_credit) +
geom_point(alpha = 0.1, position = position_jitter(h = 0)) +
geom_smooth(method = 'lm', color = 'red') +
ylab('Budget') +
xlab('Revenue') +
ggtitle('Budget Vs Revenue')
ggplot(aes(x = log(revenue), y = vote_average), data = db_credit) +
geom_point(alpha = 0.1, position = position_jitter(h = 0)) +
geom_smooth(method = 'lm', color = 'red') +
ylab('TMDB score') +
xlab('Revenue') +
ggtitle('TMDB score vs Revenue')
ggplot(aes(x = log(budget), y = vote_average), data = db_credit) +
geom_point(alpha = 0.1, position = position_jitter(h = 0)) +
geom_smooth(method = 'lm', color = 'red') +
ylab('TMDB score') +
xlab('Budget') +
ggtitle('TMDB score vs Budget')
1.6. Average Movie Rating
ggplot(db_credit,aes(vote_average)) +
geom_histogram(bins = 100) +
geom_vline(xintercept = mean(tmdb$vote_average,na.rm = TRUE),colour = "red") +
ylab("Count of Movies") +
xlab("Average Vote") +
ggtitle("Histogram for average vote rating")
Mean : 6.092083
Let us look at the top 20 movies with highest average_vote with color according to vote count.
Here, movies with vote count > 500 are considered as movies with less vote counts and high rating can be a misleading statistic.
head(db_credit)1.7. Trasformation of JSON column to unique columns to the movie dataset
## transformation of "keywords" column into tibble
keywords <- movie %>%
filter(nchar(keywords) > 2) %>% # fiter out blank keywords field
mutate( # create a new field
js = lapply(keywords, fromJSON) # containing a LIST of keyword and value pairs
) %>% # called id and name
unnest(js) %>% # turn each keyword/value pairs in the LIST into a row
dplyr::select(id, title, keywords = name)
## Combining the keywords of a movie in a single column
keywords <- aggregate(keywords ~.,data = keywords, paste, collapse = ",")
#Combining the genres of a movie in a single column
genres <- movie %>% filter(nchar(genres) > 2) %>%
mutate( js = lapply(genres, fromJSON)) %>%
unnest(js) %>%
dplyr::select(id, title, genres = name)
genres <- aggregate(genres ~.,data = genres, paste, collapse = ",")
# Combining production_companies
production_companies <- movie %>% filter(nchar(production_companies) > 2) %>%
mutate( js = lapply(production_companies, fromJSON)) %>%
unnest(js) %>%
dplyr::select(id, title, production_companies = name)
production_companies <- aggregate(production_companies ~.,data = production_companies, paste, collapse = ",")
# Combining production countries
production_countries <- movie %>%
filter(nchar(production_countries) > 2) %>%
mutate(
js = lapply(production_countries, fromJSON)
) %>%
unnest(js) %>%
dplyr::select(id, title, production_countries = name)
countries <- movie %>%
filter(nchar(production_countries) > 2) %>%
mutate(
js = lapply(production_countries, fromJSON)
) %>%
unnest(js) %>%
dplyr::select(id, title, production_countries = name)
production_countries <- aggregate(production_countries ~.,data = production_countries, paste, collapse = ",")
# combining spoken languages
spoken_languages <- movie %>%
filter(nchar(spoken_languages) > 2) %>%
mutate(
js = lapply(spoken_languages, fromJSON)
) %>%
unnest(js) %>%
dplyr::select(id, title, spoken_languages = iso_639_1)
spoken_languages <- aggregate(spoken_languages ~.,data = spoken_languages, paste, collapse = ",")movies <- subset(movie, select = -c(genres, keywords, production_companies, production_countries,spoken_languages))
glimpse(movies)Observations: 4,803
Variables: 18
$ budget <int> 237000000, 300000000, 245000000, 250000000, 260000000, 258000000, 260000000, 280000000, 250000000, 25000...
$ homepage <chr> "http://www.avatarmovie.com/", "http://disney.go.com/disneypictures/pirates/", "http://www.sonypictures....
$ id <int> 19995, 285, 206647, 49026, 49529, 559, 38757, 99861, 767, 209112, 1452, 10764, 58, 57201, 49521, 2454, 2...
$ original_language <chr> "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "e...
$ original_title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carter",...
$ overview <chr> "In the 22nd century, a paraplegic Marine is dispatched to the moon Pandora on a unique mission, but bec...
$ popularity <dbl> 150.437577, 139.082615, 107.376788, 112.312950, 43.926995, 115.699814, 48.681969, 134.279229, 98.885637,...
$ release_date <date> 2009-12-10, 2007-05-19, 2015-10-26, 2012-07-16, 2012-03-07, 2007-05-01, 2010-11-24, 2015-04-22, 2009-07...
$ revenue <dbl> 2787965087, 961000000, 880674609, 1084939099, 284139100, 890871626, 591794936, 1405403694, 933959197, 87...
$ runtime <int> 162, 169, 148, 165, 132, 139, 100, 141, 153, 151, 154, 106, 151, 149, 143, 150, 143, 136, 106, 144, 136,...
$ status <chr> "Released", "Released", "Released", "Released", "Released", "Released", "Released", "Released", "Release...
$ tagline <chr> "Enter the World of Pandora.", "At the end of the world, the adventure begins.", "A Plan No One Escapes"...
$ title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carter",...
$ vote_average <dbl> 7.2, 6.9, 6.3, 7.6, 6.1, 5.9, 7.4, 7.3, 7.4, 5.7, 5.4, 6.1, 7.0, 5.9, 6.5, 6.3, 7.4, 6.4, 6.2, 7.1, 6.5,...
$ vote_count <int> 11800, 4500, 4466, 9106, 2124, 3576, 3330, 6767, 5293, 7004, 1400, 2965, 5246, 2311, 6359, 1630, 11776, ...
$ Year <fct> 2009, 2007, 2015, 2012, 2012, 2007, 2010, 2015, 2009, 2016, 2006, 2008, 2006, 2013, 2013, 2008, 2012, 20...
$ Date <fct> 10, 19, 26, 16, 07, 01, 24, 22, 07, 23, 28, 30, 20, 03, 12, 15, 25, 14, 23, 10, 27, 12, 11, 04, 14, 18, ...
$ month <chr> "Dec", "May", "Oct", "Jul", "Mar", "May", "Nov", "Apr", "Jul", "Mar", "Jun", "Oct", "Jun", "Jul", "Jun",...# Dropped existing unformatted columns in the main dataset, creating a new dataset "movies"
movies <- subset(movie, select = -c(genres, keywords, production_companies, production_countries, spoken_languages))
movies <- movies %>%
full_join(keywords, by = c("id", "title")) %>%
full_join(genres, by = c("id", "title")) %>%
full_join(production_companies, by = c("id", "title")) %>%
full_join(production_countries, by = c("id", "title")) %>%
full_join(spoken_languages, by = c("id", "title"))
glimpse(movies)Observations: 4,803
Variables: 23
$ budget <int> 237000000, 300000000, 245000000, 250000000, 260000000, 258000000, 260000000, 280000000, 250000000, 25...
$ homepage <chr> "http://www.avatarmovie.com/", "http://disney.go.com/disneypictures/pirates/", "http://www.sonypictur...
$ id <int> 19995, 285, 206647, 49026, 49529, 559, 38757, 99861, 767, 209112, 1452, 10764, 58, 57201, 49521, 2454...
$ original_language <chr> "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en", "en",...
$ original_title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carte...
$ overview <chr> "In the 22nd century, a paraplegic Marine is dispatched to the moon Pandora on a unique mission, but ...
$ popularity <dbl> 150.437577, 139.082615, 107.376788, 112.312950, 43.926995, 115.699814, 48.681969, 134.279229, 98.8856...
$ release_date <date> 2009-12-10, 2007-05-19, 2015-10-26, 2012-07-16, 2012-03-07, 2007-05-01, 2010-11-24, 2015-04-22, 2009...
$ revenue <dbl> 2787965087, 961000000, 880674609, 1084939099, 284139100, 890871626, 591794936, 1405403694, 933959197,...
$ runtime <int> 162, 169, 148, 165, 132, 139, 100, 141, 153, 151, 154, 106, 151, 149, 143, 150, 143, 136, 106, 144, 1...
$ status <chr> "Released", "Released", "Released", "Released", "Released", "Released", "Released", "Released", "Rele...
$ tagline <chr> "Enter the World of Pandora.", "At the end of the world, the adventure begins.", "A Plan No One Escap...
$ title <chr> "Avatar", "Pirates of the Caribbean: At World's End", "Spectre", "The Dark Knight Rises", "John Carte...
$ vote_average <dbl> 7.2, 6.9, 6.3, 7.6, 6.1, 5.9, 7.4, 7.3, 7.4, 5.7, 5.4, 6.1, 7.0, 5.9, 6.5, 6.3, 7.4, 6.4, 6.2, 7.1, 6...
$ vote_count <int> 11800, 4500, 4466, 9106, 2124, 3576, 3330, 6767, 5293, 7004, 1400, 2965, 5246, 2311, 6359, 1630, 1177...
$ Year <fct> 2009, 2007, 2015, 2012, 2012, 2007, 2010, 2015, 2009, 2016, 2006, 2008, 2006, 2013, 2013, 2008, 2012,...
$ Date <fct> 10, 19, 26, 16, 07, 01, 24, 22, 07, 23, 28, 30, 20, 03, 12, 15, 25, 14, 23, 10, 27, 12, 11, 04, 14, 1...
$ month <chr> "Dec", "May", "Oct", "Jul", "Mar", "May", "Nov", "Apr", "Jul", "Mar", "Jun", "Oct", "Jun", "Jul", "Ju...
$ keywords <chr> "culture clash,future,space war,space colony,society,space travel,futuristic,romance,space,alien,trib...
$ genres <chr> "Action,Adventure,Fantasy,Science Fiction", "Adventure,Fantasy,Action", "Action,Adventure,Crime", "Ac...
$ production_companies <chr> "Ingenious Film Partners,Twentieth Century Fox Film Corporation,Dune Entertainment,Lightstorm Enterta...
$ production_countries <chr> "United States of America,United Kingdom", "United States of America", "United Kingdom,United States ...
$ spoken_languages <chr> "en,es", "en", "fr,en,es,it,de", "en", "en", "en,fr", "en", "en", "en", "en", "en,fr,de", "en,es,it,f...1.8. Movie Rating Vs Budget
movies %>% dplyr::select(title,vote_average,vote_count, budget) %>%
filter(vote_count > 500 ) %>% arrange(desc(vote_average)) %>% head(20) %>%
ggplot(aes(x = title,y = vote_average,fill = budget )) + geom_bar(stat = "identity") + coord_flip(ylim = c(7, 9)) +
scale_fill_continuous()
1.9. Popularity vs Budget are not highly correlated
#Top 20 movies by popularity, color according to vote count :
movies %>% dplyr::select(title,vote_average,vote_count, popularity) %>%
filter(vote_count > 300 ) %>% head(30) %>%
ggplot(aes(x = title,y = popularity, fill = vote_count)) + geom_bar(stat = "identity") + coord_flip() +
scale_fill_continuous()
1.10. Analysing Movie Genre
genre1 <- Corpus(VectorSource(genres$genres))
dtm <- DocumentTermMatrix(genre1)
genre_freq <- colSums(as.matrix(dtm))
freq <- sort(colSums(as.matrix(dtm)), decreasing = TRUE)
genre_wframe <- data.frame(word = names(genre_freq), freq = genre_freq)
ggplot(genre_wframe, aes(x = reorder(word,-freq), y = freq)) +
geom_bar(stat = "identity") +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
ggtitle("Movie Genre frequency graph") +
xlab("Genre") +
ylab("Frequency")
1.11. Compare Profit with the movie rating and budget
#Adding new columns gross and gross_flag
movies <- movies %>%
mutate(gross = revenue - budget, gross_flag = ifelse(gross < 0, "Loss", "Profit"))
library(plotly)
Attaching package: ‘plotly’
The following object is masked from ‘package:MASS’:
select
The following object is masked from ‘package:formattable’:
style
The following object is masked from ‘package:ggplot2’:
last_plot
The following objects are masked from ‘package:plyr’:
arrange, mutate, rename, summarise
The following object is masked from ‘package:stats’:
filter
The following object is masked from ‘package:graphics’:
layoutplot_ly(movies, x = ~vote_average, y = ~budget, z = ~gross/1000000,
color = ~gross_flag, colors = c('#BF382A', '#0C4B8E') ,size = I(3)) %>%
add_markers() %>%
layout(scene = list(xaxis = list(title = 'Average vote'),
yaxis = list(title = 'Budget'),
zaxis = list(title = 'Gross (million $)')),
title = "INTERACTIVE 3D Scatter plot: Average vote vs Budget vs Gross",
showlegend = FALSE)The blue dots represents the movie with movies making profits, we can hover over the plot to undersatnd the behaviour of the variables vote_average anf budget with respect to Gross.
2. Data Cleaning
movies <- read.csv("tmdb_5000_movies.csv", header = TRUE, stringsAsFactors = FALSE)
credits <- read.csv("tmdb_5000_credits.csv",header = TRUE, stringsAsFactors = FALSE )library(tidyverse)
library(jsonlite)#movies <- tmdb_5000_movies
#credits <- tmdb_5000_credits
Top.100.Directors.Working.Today <- read.csv("Top 100 Directors Working Today.csv", header = TRUE, stringsAsFactors = FALSE)
The.Top.100.Actors.of.2016 <- read.csv("The Top 100 Actors of 2016.csv",header = TRUE, stringsAsFactors = FALSE )
top100directors <- Top.100.Directors.Working.Today$Name
top100directors <- paste(top100directors)
top100actors <- The.Top.100.Actors.of.2016$Name
top100actors <- paste(top100actors)
# from: http://www.imdb.com/list/ls072596173/
#top100directors <- Top.100.Directors.Working.Today$Name
#top100directors <- paste(top100directors)
#top100actors <- The.Top.100.Actors.of.2016$Name
#top100actors <- paste(top100actors)# Drop useless columns
movies <- movies[,!(names(movies) %in% c('homepage','overview','status',
'title','tagline','original_title'))]
credits <- credits[,!(names(credits) %in% c('title'))]
dim(movies)[1] 4803 142.1. Let’s clean movies dataset
2.1.1. Distinct Months; Holiday month
# Extract month of release date; put into new column
movies$release_date <- as.POSIXlt(movies$release_date, format="%Y-%m-%d")
movies$release_month <- lubridate::month(movies$release_date)
# From Linear Regression using each of 12 months as dummy categories,
# we saw that months 5, 6, 11 and 12 are important,
# while the rest are unimportant.
# We will aggregate this as 'holiday month' (beginning of Summer; beginning of Winter)
movies$holiday_month <- 0
movies[which(movies$release_month %in% c(5,6,11,12)),]['holiday_month'] <- 1
#drop release date and month
movies <- movies[ , !(names(movies) %in% c('release_date','release_month'))]2.1.2. Number of Languages; English / No english
# count number of languages
movies$num_lang <- stringr::str_count(movies$spoken_languages, "\"name\":")
movies$original_isEnglish <- 0
movies[which(movies$original_language == 'en'),]['original_isEnglish'] <- 1
# drop original lanugage; spoken languages
movies <- movies[ , !(names(movies) %in% c('original_language', 'spoken_languages'))]2.1.3. Distinct Production Countries
# turn JSON into DF with one country per row
countryDF = movies %>% filter(nchar(as.character(production_countries))>2) %>% mutate(js=lapply(as.character(production_countries),fromJSON)) %>% unnest(js) %>% dplyr::select(id, production_countries=name)
# Select Unique Countries
# Distinct_Countries = countryDF %>% distinct(production_countries)
df <- as.data.frame(table(countryDF$production_countries))
head(with(df,df[order(Freq,decreasing = TRUE),]))We will make dummy variables on top 6 frequent countries on the list.
# United States
movies$country_USA <- 0
movies$country_UK <- 0
movies$country_Germany <- 0
movies$country_France <- 0
movies$country_Canada <- 0
movies$country_Austr <- 0
for (i in 1:nrow(movies)) {
noQuotes <- strsplit(toString(movies$production_countries[i]),'\"')
noQuotes <- unlist(noQuotes)
if (any(noQuotes == "United States of America")) {
movies$country_USA[i] <- 1
}
if (any(noQuotes == "United Kingdom")) {
movies$country_UK[i] <- 1
}
if (any(noQuotes == "Germany")) {
movies$country_Germany[i] <- 1
}
if (any(noQuotes == "France")) {
movies$country_France[i] <- 1
}
if (any(noQuotes == "Canada")) {
movies$country_Canada[i] <- 1
}
if (any(noQuotes == "Australia")) {
movies$country_Austr[i] <- 1
}
}
# drop production countries column
movies <- movies[,!(names(movies) %in% c('production_countries'))]# Check correct numbers below.
# Should match with frequencies from
# head(with(df,df[order(Freq,decreasing = TRUE),]))
as.data.frame(table(movies$country_UK))2.1.4. Distinct Genres
# turn JSON into DF with one genre per row
genreDF= movies %>% filter(nchar(as.character(genres))>2) %>% mutate(js=lapply(as.character(genres),fromJSON)) %>% unnest(js) %>% dplyr::select(id, genre=name)
# slice(genredf) # View the newly created DF
# Select Unique Genres
Distinct_Genres = unique(genreDF$genre)
# Make dummy variable for each genre
for (name in Distinct_Genres) {
# prepend 'genre' to all genre names
col_Name <- paste('genre', name, sep='_')
movies[col_Name] <- 0
movies[grep(name,movies$genres),][col_Name] <- 1
}
# Drop raw json genres
movies <- movies[,!(names(movies) %in% c('genres'))]2.1.5. Distinct Keywords
# turn JSON into DF with one keyword per row
keywordDF=movies %>% filter(nchar(as.character(keywords))>2) %>% mutate(js=lapply(as.character(keywords),fromJSON)) %>% unnest(js) %>% dplyr::select(id,keywords=name)
# Select Unique Genres
# Distinct_Keywords = unique(keywordDF$keywords)
# length(Distinct_Keywords)
df <- as.data.frame(table(keywordDF$keywords))
with(df,df[order(Freq,decreasing = TRUE),])2.1.6. Distinct Production Companies
# turn JSON into DF with one company per row
companyDF=movies %>% filter(nchar(as.character(production_companies))>2) %>% mutate(js=lapply(as.character(production_companies),fromJSON)) %>% unnest(js) %>% dplyr::select(id,production_companies=name)
# Select Unique Genres
# Distinct_Companies = companyDF %>% distinct(production_companies)
df <- as.data.frame(table(companyDF$production_companies))
with(df,df[order(Freq,decreasing = TRUE),])2.2. Lets Clean Credits Dataset
# Split JSON for cast
all_cast <- credits %>% # start with the raw tibble
filter(nchar(as.character(cast)) > 2) %>% # filter out movies with empty cast
mutate( #
js_cast = lapply(as.character(cast), fromJSON) # turn the JSON into a list
) %>% #
unnest(js_cast) %>% dplyr::select(movie_id, cast_id, character, credit_id, gender, id, name = name, order)
# Split JSON for crew
all_crew <- credits %>% # start with the raw tibble
filter(nchar(as.character(crew)) > 2) %>% # filter out movies with empty crew
mutate(
js_crew = lapply(as.character(crew), fromJSON) # turn the JSON into a list
) %>% #
unnest(js_crew) %>% dplyr::select(movie_id, credit_id, department, gender, id, job, name = name)2.2.1. Check if Director is in top 100
# List all movie directors
directors <- filter(all_crew[,c('movie_id','job','name')], all_crew$job == 'Director')
directors# List of top 100 directors
with(Top.100.Directors.Working.Today, Top.100.Directors.Working.Today[order(Name),])['Name']Make a column to represent if the movie has top director.
# New binary column to aggregate by director fame
credits$topDirector <- 0
# If director in top 100 directors list, change from 0 to 1
for (i in 1:nrow(directors)) {
if (directors$name[i] %in% top100directors) { # match director and credits DFs by ID
credits[which(credits$movie_id == directors$movie_id[i]),]['topDirector'] <- 1
}
}
# Drop crew column
credits <- credits[,!(names(credits) %in% c('crew'))]
credits# Check value frequencies
as.data.frame(table(credits$topDirector))2.2.2. Check if any top 100 Actors
# List all movie directors
actors <- filter(all_cast[,c('movie_id','name')])
actorsMake a column to represent if the movie has top actor.
# New binary column to aggregate by director fame
credits$topActor <- 0
# If Actors in top 100 actors list, change from 0 to 1
for (i in 1:nrow(actors)) {
if (actors$name[i] %in% top100actors) { # match actors and credits DFs by ID
credits[which(credits$movie_id == actors$movie_id[i]),]['topActor'] <- 1
}
}
head(credits$topActor)[1] 0 1 1 1 1 12.2.3. Average Genders of Cast
# 0 = unknown
# 1 = Female
# 2 = Male
CastGenders <- filter(all_cast[,c('movie_id','gender')],
all_cast$gender != 0)
CastGenders <- aggregate(CastGenders$gender, by=list(CastGenders$movie_id), FUN=mean)
# Append Avg Genders to Credits
credits$CastGenderAVG <- 1.5
for (ID in CastGenders$Group.1){
credits[which(credits$movie_id == ID),]['CastGenderAVG'] <-
CastGenders[which(CastGenders$Group.1 == ID),]['x']
}
# Drop cast column
credits <- credits[,!(names(credits) %in% c('cast'))]2.2.4. Merge movie and credit dataset
TMDB_Cleaned <- left_join(credits,movies, by = c('movie_id' = 'id'))
colnames(TMDB_Cleaned) [1] "movie_id" "topDirector" "topActor" "CastGenderAVG" "budget"
[6] "keywords" "popularity" "production_companies" "revenue" "runtime"
[11] "vote_average" "vote_count" "holiday_month" "num_lang" "original_isEnglish"
[16] "country_USA" "country_UK" "country_Germany" "country_France" "country_Canada"
[21] "country_Austr" "genre_Action" "genre_Adventure" "genre_Fantasy" "genre_Science Fiction"
[26] "genre_Crime" "genre_Drama" "genre_Thriller" "genre_Animation" "genre_Family"
[31] "genre_Western" "genre_Comedy" "genre_Romance" "genre_Horror" "genre_Mystery"
[36] "genre_History" "genre_War" "genre_Music" "genre_Documentary" "genre_Foreign"
[41] "genre_TV Movie" dim(TMDB_Cleaned)[1] 4803 412.2.5. Remove keywords and production_companies from the dataset as they are huge and is less significant to predict the revenue
# remove this after cleaning keywords and companies properly
TMDB_Cleaned <- TMDB_Cleaned[,!(names(TMDB_Cleaned) %in% c('keywords','production_companies'))]
TMDB_Cleaned <- na.omit(TMDB_Cleaned)
dim(TMDB_Cleaned)[1] 4801 39write.csv(TMDB_Cleaned, file='TMDB_Cleaned.csv')tmdb <- read.csv("TMDB_Cleaned.csv", header = TRUE, stringsAsFactors = FALSE)
tmdb1<-subset(tmdb,select=-c(X,movie_id))
head(tmdb1)2.2.6. Concise num_lang to 4 spoken languages in a movie for better analysis
tmdb1 <- tmdb1 %>% mutate(num_lang = ifelse(num_lang >= 5, 4, num_lang))
df <- as.data.frame(table(tmdb1$num_lang))
head(with(df,df[order(Freq,decreasing = TRUE),]))head(tmdb1)2.2.7. Converting binary columns to categorical variables
tmdb1$original_isEnglish<- as.factor(tmdb1$original_isEnglish)
tmdb1$holiday_month<- as.factor(tmdb1$holiday_month)
tmdb1$topDirector<- as.factor(tmdb1$topDirector)
tmdb1$topActor <- as.factor(tmdb1$topActor)
tmdb1$num_lang <- as.factor(tmdb1$num_lang)
tmdb1$genre_Crime <- as.factor(tmdb1$genre_Crime)
tmdb1$genre_Science.Fiction <- as.factor(tmdb1$genre_Science.Fiction)
tmdb1$genre_Family<- as.factor(tmdb1$genre_Family)
tmdb1$genre_Animation <- as.factor(tmdb1$genre_Animation)
tmdb1$genre_Western <- as.factor(tmdb1$genre_Western)
tmdb1$genre_Adventure <- as.factor(tmdb1$genre_Adventure)
tmdb1$genre_Romance <- as.factor(tmdb1$genre_Romance)
tmdb1$genre_Drama <- as.factor(tmdb1$genre_Drama)
tmdb1$genre_Action <- as.factor(tmdb1$genre_Action)
tmdb1$genre_Documentary <- as.factor(tmdb1$genre_Documentary)
tmdb1$genre_Comedy <- as.factor(tmdb1$genre_Comedy)
tmdb1$genre_Horror <- as.factor(tmdb1$genre_Horror)
tmdb1$genre_Mystery <- as.factor(tmdb1$genre_Mystery)
tmdb1$genre_History <- as.factor(tmdb1$genre_History)
tmdb1$genre_Music <- as.factor(tmdb1$genre_History)
tmdb1$genre_Foreign <- as.factor(tmdb1$genre_Foreign)
tmdb1$genre_TV.Movie <- as.factor(tmdb1$genre_TV.Movie)
tmdb1$genre_Fantasy <- as.factor(tmdb1$genre_Fantasy)
tmdb1$genre_Thriller <- as.factor(tmdb1$genre_Thriller)
tmdb1$genre_War <- as.factor(tmdb1$genre_War)
tmdb1$country_USA <- as.factor(tmdb1$country_USA)
tmdb1$country_UK <- as.factor(tmdb1$country_UK)
tmdb1$country_France <- as.factor(tmdb1$country_France)
tmdb1$country_Germany <- as.factor(tmdb1$country_Germany)
tmdb1$country_Canada <- as.factor(tmdb1$country_Canada)
tmdb1$country_Austr <- as.factor(tmdb1$country_Austr)
#tmdb2$revenue <- log(tmdb2$revenue)
#tmdb2$budget <- log(tmdb2$budget)
str(tmdb1)'data.frame': 4801 obs. of 38 variables:
$ topDirector : Factor w/ 2 levels "0","1": 2 1 1 2 1 1 1 2 1 1 ...
$ topActor : Factor w/ 2 levels "0","1": 1 2 2 2 2 2 1 2 1 2 ...
$ CastGenderAVG : num 1.62 1.83 1.81 1.87 1.76 ...
$ budget : int 237000000 300000000 245000000 250000000 260000000 258000000 260000000 280000000 250000000 250000000 ...
$ popularity : num 150.4 139.1 107.4 112.3 43.9 ...
$ revenue : num 2.79e+09 9.61e+08 8.81e+08 1.08e+09 2.84e+08 ...
$ runtime : int 162 169 148 165 132 139 100 141 153 151 ...
$ vote_average : num 7.2 6.9 6.3 7.6 6.1 5.9 7.4 7.3 7.4 5.7 ...
$ vote_count : int 11800 4500 4466 9106 2124 3576 3330 6767 5293 7004 ...
$ holiday_month : Factor w/ 2 levels "0","1": 2 2 1 1 1 2 2 1 1 1 ...
$ num_lang : Factor w/ 5 levels "0","1","2","3",..: 3 2 5 2 2 3 2 2 2 2 ...
$ original_isEnglish : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 2 2 2 2 ...
$ country_USA : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 2 2 2 2 ...
$ country_UK : Factor w/ 2 levels "0","1": 2 1 2 1 1 1 1 1 2 1 ...
$ country_Germany : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ country_France : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ country_Canada : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ country_Austr : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Action : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 1 2 1 2 ...
$ genre_Adventure : Factor w/ 2 levels "0","1": 2 2 2 1 2 2 1 2 2 2 ...
$ genre_Fantasy : Factor w/ 2 levels "0","1": 2 2 1 1 1 2 1 1 2 2 ...
$ genre_Science.Fiction: Factor w/ 2 levels "0","1": 2 1 1 1 2 1 1 2 1 1 ...
$ genre_Crime : Factor w/ 2 levels "0","1": 1 1 2 2 1 1 1 1 1 1 ...
$ genre_Drama : Factor w/ 2 levels "0","1": 1 1 1 2 1 1 1 1 1 1 ...
$ genre_Thriller : Factor w/ 2 levels "0","1": 1 1 1 2 1 1 1 1 1 1 ...
$ genre_Animation : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 2 1 1 1 ...
$ genre_Family : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 2 1 2 1 ...
$ genre_Western : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Comedy : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Romance : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Horror : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Mystery : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_History : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_War : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Music : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Documentary : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Foreign : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_TV.Movie : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...2.2.8. Scale the data for uniformity in numerical values of the dataset
library(MASS)
ind <- sapply(tmdb1, is.numeric)
tmdb1[ind] <- lapply(tmdb1[ind], scale)
str(tmdb1)'data.frame': 4801 obs. of 38 variables:
$ topDirector : Factor w/ 2 levels "0","1": 2 1 1 2 1 1 1 2 1 1 ...
$ topActor : Factor w/ 2 levels "0","1": 1 2 2 2 2 2 1 2 1 2 ...
$ CastGenderAVG : num [1:4801, 1] -0.215 0.946 0.81 1.145 0.538 ...
..- attr(*, "scaled:center")= num 1.66
..- attr(*, "scaled:scale")= num 0.179
$ budget : num [1:4801, 1] 5.11 6.65 5.3 5.42 5.67 ...
..- attr(*, "scaled:center")= num 29054015
..- attr(*, "scaled:scale")= num 40728211
$ popularity : num [1:4801, 1] 4.052 3.695 2.699 2.854 0.705 ...
..- attr(*, "scaled:center")= num 21.5
..- attr(*, "scaled:scale")= num 31.8
$ revenue : num [1:4801, 1] 16.61 5.39 4.9 6.16 1.24 ...
..- attr(*, "scaled:center")= num 82294907
..- attr(*, "scaled:scale")= num 1.63e+08
$ runtime : num [1:4801, 1] 2.44 2.75 1.82 2.57 1.11 ...
..- attr(*, "scaled:center")= num 107
..- attr(*, "scaled:scale")= num 22.6
$ vote_average : num [1:4801, 1] 0.92893 0.67714 0.17357 1.26464 0.00572 ...
..- attr(*, "scaled:center")= num 6.09
..- attr(*, "scaled:scale")= num 1.19
$ vote_count : num [1:4801, 1] 9 3.09 3.06 6.82 1.16 ...
..- attr(*, "scaled:center")= num 691
..- attr(*, "scaled:scale")= num 1235
$ holiday_month : Factor w/ 2 levels "0","1": 2 2 1 1 1 2 2 1 1 1 ...
$ num_lang : Factor w/ 5 levels "0","1","2","3",..: 3 2 5 2 2 3 2 2 2 2 ...
$ original_isEnglish : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 2 2 2 2 ...
$ country_USA : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 2 2 2 2 ...
$ country_UK : Factor w/ 2 levels "0","1": 2 1 2 1 1 1 1 1 2 1 ...
$ country_Germany : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ country_France : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ country_Canada : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ country_Austr : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Action : Factor w/ 2 levels "0","1": 2 2 2 2 2 2 1 2 1 2 ...
$ genre_Adventure : Factor w/ 2 levels "0","1": 2 2 2 1 2 2 1 2 2 2 ...
$ genre_Fantasy : Factor w/ 2 levels "0","1": 2 2 1 1 1 2 1 1 2 2 ...
$ genre_Science.Fiction: Factor w/ 2 levels "0","1": 2 1 1 1 2 1 1 2 1 1 ...
$ genre_Crime : Factor w/ 2 levels "0","1": 1 1 2 2 1 1 1 1 1 1 ...
$ genre_Drama : Factor w/ 2 levels "0","1": 1 1 1 2 1 1 1 1 1 1 ...
$ genre_Thriller : Factor w/ 2 levels "0","1": 1 1 1 2 1 1 1 1 1 1 ...
$ genre_Animation : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 2 1 1 1 ...
$ genre_Family : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 2 1 2 1 ...
$ genre_Western : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Comedy : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Romance : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Horror : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Mystery : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_History : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_War : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Music : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Documentary : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_Foreign : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
$ genre_TV.Movie : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...2.2.9. Let’s explore data to look at our numerical values
summary(tmdb1) topDirector topActor CastGenderAVG.V1 budget.V1 popularity.V1 revenue.V1 runtime.V1
0:4397 0:3296 Min. :-3.696618 Min. :-0.713363 Min. :-0.675702 Min. :-0.505241 Min. :-4.726524
1: 404 1:1505 1st Qu.:-0.657740 1st Qu.:-0.693721 1st Qu.:-0.528620 1st Qu.:-0.505241 1st Qu.:-0.569428
Median : 0.017566 Median :-0.345068 Median :-0.269413 Median :-0.387488 Median :-0.171408
Mean : 0.000000 Mean : 0.000000 Mean : 0.000000 Mean : 0.000000 Mean : 0.000000
3rd Qu.: 0.692872 3rd Qu.: 0.268757 3rd Qu.: 0.215253 3rd Qu.: 0.065239 3rd Qu.: 0.491959
Max. : 1.874658 Max. : 8.616779 Max. :26.840682 Max. :16.611191 Max. :10.221334
vote_average.V1 vote_count.V1 holiday_month num_lang original_isEnglish country_USA country_UK country_Germany
Min. :-5.113911 Min. :-0.559219 0:3269 0: 85 0: 297 0: 845 0:4166 0:4477
1st Qu.:-0.413925 1st Qu.:-0.515486 1:1532 1:3354 1:4504 1:3956 1: 635 1: 324
Median : 0.089645 Median :-0.368089 2: 841
Mean : 0.000000 Mean : 0.000000 3: 308
3rd Qu.: 0.593215 3rd Qu.: 0.037657 4: 213
Max. : 3.278921 Max. :10.578132
country_France country_Canada country_Austr genre_Action genre_Adventure genre_Fantasy genre_Science.Fiction genre_Crime genre_Drama
0:4495 0:4540 0:4691 0:3647 0:4011 0:4377 0:4266 0:4105 0:2505
1: 306 1: 261 1: 110 1:1154 1: 790 1: 424 1: 535 1: 696 1:2296
genre_Thriller genre_Animation genre_Family genre_Western genre_Comedy genre_Romance genre_Horror genre_Mystery genre_History
0:3527 0:4567 0:4288 0:4719 0:3079 0:3907 0:4282 0:4453 0:4604
1:1274 1: 234 1: 513 1: 82 1:1722 1: 894 1: 519 1: 348 1: 197
genre_War genre_Music genre_Documentary genre_Foreign genre_TV.Movie
0:4657 0:4604 0:4692 0:4767 0:4793
1: 144 1: 197 1: 109 1: 34 1: 8
2.2.10. Investigate for missing values
sapply(tmdb1,function(x) sum(is.na(x))) # number of missing values for each variable topDirector topActor CastGenderAVG budget popularity revenue
0 0 0 0 0 0
runtime vote_average vote_count holiday_month num_lang original_isEnglish
0 0 0 0 0 0
country_USA country_UK country_Germany country_France country_Canada country_Austr
0 0 0 0 0 0
genre_Action genre_Adventure genre_Fantasy genre_Science.Fiction genre_Crime genre_Drama
0 0 0 0 0 0
genre_Thriller genre_Animation genre_Family genre_Western genre_Comedy genre_Romance
0 0 0 0 0 0
genre_Horror genre_Mystery genre_History genre_War genre_Music genre_Documentary
0 0 0 0 0 0
genre_Foreign genre_TV.Movie
0 0 3. Predictive modeling
Models used to build the predictive model- Linear regression Lasso regression Ridge Regression Regression Trees *Random Forest
Plot scatterplot on numerical variables
scatterplotMatrix(~tmdb1$revenue+tmdb1$budget+tmdb1$popularity+tmdb1$runtime+tmdb1$CastGenderAVG+tmdb1$vote_average+tmdb1$vote_count)
library(corrplot)corrplot 0.84 loadednumeric_col <- sapply(tmdb1, is.numeric)
tmdb1_numeric<- tmdb1[,numeric_col]
Correlation<-cor(tmdb1_numeric)
corrplot(Correlation, method = "color")
3.1. Linear Regression
3.1.1. Perfrom stepwise regression to identify the top predictors
null_model<-lm(revenue~1,data=tmdb1)
full_model<-lm(revenue~.,data=tmdb1)
step(null_model, scope = list(lower = null_model, upper = full_model), direction = "both")Start: AIC=1
revenue ~ 1
Df Sum of Sq RSS AIC
+ vote_count 1 2931.28 1868.7 -4526.1
+ budget 1 2563.52 2236.5 -3663.6
+ popularity 1 1994.92 2805.1 -2576.0
+ genre_Adventure 1 569.13 4230.9 -602.9
+ runtime 1 302.63 4497.4 -309.7
+ topActor 1 255.71 4544.3 -259.8
+ genre_Fantasy 1 216.21 4583.8 -218.3
+ holiday_month 1 200.96 4599.0 -202.3
+ genre_Action 1 198.77 4601.2 -200.0
+ genre_Animation 1 190.66 4609.3 -191.6
+ vote_average 1 186.82 4613.2 -187.6
+ country_USA 1 160.60 4639.4 -160.4
+ genre_Drama 1 150.83 4649.2 -150.3
+ topDirector 1 139.81 4660.2 -138.9
+ genre_Family 1 138.73 4661.3 -137.8
+ genre_Science.Fiction 1 111.72 4688.3 -110.1
+ original_isEnglish 1 48.48 4751.5 -45.7
+ num_lang 4 52.80 4747.2 -44.1
+ CastGenderAVG 1 39.29 4760.7 -36.5
+ genre_Horror 1 32.93 4767.1 -30.1
+ genre_Documentary 1 22.01 4778.0 -19.1
+ genre_Romance 1 20.58 4779.4 -17.6
+ country_France 1 15.90 4784.1 -12.9
+ genre_Comedy 1 12.26 4787.7 -9.3
+ genre_Foreign 1 8.66 4791.3 -5.7
+ genre_Crime 1 8.00 4792.0 -5.0
+ country_Canada 1 6.43 4793.6 -3.4
+ genre_History 1 4.75 4795.2 -1.8
+ genre_Music 1 4.75 4795.2 -1.8
+ country_UK 1 4.19 4795.8 -1.2
+ genre_Western 1 4.09 4795.9 -1.1
+ genre_TV.Movie 1 2.05 4798.0 1.0
<none> 4800.0 1.0
+ genre_Mystery 1 0.23 4799.8 2.8
+ country_Germany 1 0.17 4799.8 2.8
+ country_Austr 1 0.13 4799.9 2.9
+ genre_Thriller 1 0.10 4799.9 2.9
+ genre_War 1 0.02 4800.0 3.0
Step: AIC=-4526.08
revenue ~ vote_count
Df Sum of Sq RSS AIC
+ budget 1 529.04 1339.7 -6122.0
+ genre_Adventure 1 96.14 1772.6 -4777.7
+ genre_Family 1 73.64 1795.1 -4717.1
+ genre_Animation 1 69.48 1799.2 -4706.0
+ holiday_month 1 53.52 1815.2 -4663.6
+ genre_Fantasy 1 37.66 1831.1 -4621.8
+ genre_Drama 1 32.39 1836.3 -4608.0
+ topDirector 1 16.94 1851.8 -4567.8
+ popularity 1 16.32 1852.4 -4566.2
+ genre_Crime 1 15.60 1853.1 -4564.3
+ country_USA 1 12.24 1856.5 -4555.6
+ vote_average 1 12.08 1856.6 -4555.2
+ genre_Action 1 10.11 1858.6 -4550.1
+ runtime 1 7.71 1861.0 -4543.9
+ genre_Horror 1 7.63 1861.1 -4543.7
+ genre_Thriller 1 6.21 1862.5 -4540.1
+ country_France 1 5.25 1863.5 -4537.6
+ original_isEnglish 1 4.30 1864.4 -4535.1
+ genre_Comedy 1 3.73 1865.0 -4533.7
+ genre_Mystery 1 3.27 1865.5 -4532.5
+ genre_Western 1 2.38 1866.3 -4530.2
+ country_Canada 1 2.08 1866.6 -4529.4
+ country_Germany 1 0.87 1867.9 -4526.3
+ country_UK 1 0.82 1867.9 -4526.2
<none> 1868.7 -4526.1
+ genre_Science.Fiction 1 0.75 1868.0 -4526.0
+ CastGenderAVG 1 0.47 1868.2 -4525.3
+ topActor 1 0.44 1868.3 -4525.2
+ genre_TV.Movie 1 0.32 1868.4 -4524.9
+ genre_Documentary 1 0.31 1868.4 -4524.9
+ genre_Foreign 1 0.18 1868.5 -4524.5
+ genre_Romance 1 0.17 1868.5 -4524.5
+ genre_War 1 0.14 1868.6 -4524.4
+ genre_History 1 0.11 1868.6 -4524.4
+ genre_Music 1 0.11 1868.6 -4524.4
+ country_Austr 1 0.00 1868.7 -4524.1
+ num_lang 4 1.86 1866.9 -4522.9
- vote_count 1 2931.28 4800.0 1.0
Step: AIC=-6121.97
revenue ~ vote_count + budget
Df Sum of Sq RSS AIC
+ genre_Family 1 19.34 1320.3 -6189.8
+ holiday_month 1 17.76 1321.9 -6184.0
+ genre_Animation 1 17.39 1322.3 -6182.7
+ topActor 1 12.23 1327.5 -6164.0
+ topDirector 1 11.35 1328.3 -6160.8
+ genre_Crime 1 10.58 1329.1 -6158.0
+ genre_Thriller 1 9.26 1330.4 -6153.3
+ genre_Science.Fiction 1 9.13 1330.5 -6152.8
+ genre_Adventure 1 6.65 1333.0 -6143.9
+ genre_Action 1 5.91 1333.8 -6141.2
+ country_Germany 1 5.73 1334.0 -6140.6
+ genre_Comedy 1 4.83 1334.8 -6137.3
+ genre_Drama 1 4.59 1335.1 -6136.5
+ popularity 1 4.22 1335.5 -6135.1
+ CastGenderAVG 1 3.30 1336.4 -6131.8
+ genre_Romance 1 3.27 1336.4 -6131.7
+ genre_Mystery 1 2.97 1336.7 -6130.6
+ country_France 1 2.44 1337.2 -6128.7
+ genre_Western 1 2.28 1337.4 -6128.2
+ genre_Fantasy 1 1.63 1338.0 -6125.8
+ country_Austr 1 1.09 1338.6 -6123.9
+ genre_History 1 1.07 1338.6 -6123.8
+ genre_Music 1 1.07 1338.6 -6123.8
+ genre_War 1 0.96 1338.7 -6123.4
+ country_Canada 1 0.96 1338.7 -6123.4
+ genre_Documentary 1 0.95 1338.7 -6123.4
+ country_UK 1 0.59 1339.1 -6122.1
<none> 1339.7 -6122.0
+ vote_average 1 0.46 1339.2 -6121.6
+ genre_Foreign 1 0.22 1339.5 -6120.8
+ runtime 1 0.20 1339.5 -6120.7
+ country_USA 1 0.17 1339.5 -6120.6
+ genre_Horror 1 0.02 1339.7 -6120.1
+ original_isEnglish 1 0.01 1339.7 -6120.0
+ genre_TV.Movie 1 0.00 1339.7 -6120.0
+ num_lang 4 1.52 1338.2 -6119.4
- budget 1 529.04 1868.7 -4526.1
- vote_count 1 896.80 2236.5 -3663.6
Step: AIC=-6189.8
revenue ~ vote_count + budget + genre_Family
Df Sum of Sq RSS AIC
+ holiday_month 1 16.80 1303.5 -6249.3
+ topDirector 1 9.77 1310.6 -6223.4
+ topActor 1 9.46 1310.9 -6222.3
+ genre_Science.Fiction 1 8.38 1312.0 -6218.4
+ genre_Crime 1 7.42 1312.9 -6214.8
+ genre_Animation 1 4.97 1315.4 -6205.9
+ country_Germany 1 4.79 1315.5 -6205.3
+ genre_Thriller 1 4.73 1315.6 -6205.0
+ genre_Romance 1 4.32 1316.0 -6203.5
+ popularity 1 3.86 1316.5 -6201.8
+ genre_Adventure 1 3.37 1317.0 -6200.1
+ CastGenderAVG 1 3.01 1317.3 -6198.8
+ genre_Action 1 3.01 1317.3 -6198.8
+ genre_Drama 1 2.41 1317.9 -6196.6
+ genre_Comedy 1 2.12 1318.2 -6195.5
+ country_France 1 2.11 1318.2 -6195.5
+ genre_Western 1 1.91 1318.4 -6194.7
+ genre_Mystery 1 1.85 1318.5 -6194.5
+ country_Austr 1 1.18 1319.2 -6192.1
+ genre_Documentary 1 1.08 1319.3 -6191.7
+ country_Canada 1 0.75 1319.6 -6190.5
<none> 1320.3 -6189.8
+ genre_History 1 0.51 1319.8 -6189.7
+ genre_Music 1 0.51 1319.8 -6189.7
+ genre_War 1 0.47 1319.9 -6189.5
+ country_UK 1 0.42 1319.9 -6189.3
+ vote_average 1 0.36 1320.0 -6189.1
+ genre_Foreign 1 0.20 1320.1 -6188.5
+ runtime 1 0.20 1320.1 -6188.5
+ country_USA 1 0.18 1320.2 -6188.4
+ genre_Fantasy 1 0.13 1320.2 -6188.3
+ genre_Horror 1 0.07 1320.3 -6188.1
+ genre_TV.Movie 1 0.05 1320.3 -6188.0
+ original_isEnglish 1 0.02 1320.3 -6187.9
+ num_lang 4 1.21 1319.1 -6186.2
- genre_Family 1 19.34 1339.7 -6122.0
- budget 1 474.75 1795.1 -4717.1
- vote_count 1 909.88 2230.2 -3675.1
Step: AIC=-6249.27
revenue ~ vote_count + budget + genre_Family + holiday_month
Df Sum of Sq RSS AIC
+ topDirector 1 10.42 1293.1 -6285.8
+ topActor 1 10.27 1293.3 -6285.2
+ genre_Science.Fiction 1 8.21 1295.3 -6277.6
+ genre_Crime 1 6.52 1297.0 -6271.3
+ genre_Animation 1 4.66 1298.9 -6264.5
+ country_Germany 1 4.04 1299.5 -6262.2
+ genre_Romance 1 3.89 1299.7 -6261.6
+ genre_Thriller 1 3.64 1299.9 -6260.7
+ popularity 1 3.58 1300.0 -6260.5
+ CastGenderAVG 1 3.35 1300.2 -6259.6
+ genre_Drama 1 3.27 1300.3 -6259.3
+ genre_Adventure 1 3.18 1300.4 -6259.0
+ genre_Action 1 2.70 1300.8 -6257.2
+ genre_Comedy 1 2.35 1301.2 -6255.9
+ genre_Western 1 2.30 1301.2 -6255.8
+ country_France 1 2.20 1301.3 -6255.4
+ genre_Mystery 1 1.24 1302.3 -6251.8
+ genre_Documentary 1 1.14 1302.4 -6251.5
+ country_Austr 1 1.06 1302.5 -6251.2
+ genre_History 1 0.87 1302.7 -6250.5
+ genre_Music 1 0.87 1302.7 -6250.5
+ vote_average 1 0.77 1302.8 -6250.1
+ genre_War 1 0.72 1302.8 -6249.9
+ country_Canada 1 0.58 1303.0 -6249.4
<none> 1303.5 -6249.3
+ country_UK 1 0.52 1303.0 -6249.2
+ genre_Foreign 1 0.28 1303.2 -6248.3
+ genre_Horror 1 0.27 1303.3 -6248.3
+ country_USA 1 0.16 1303.4 -6247.9
+ genre_Fantasy 1 0.08 1303.5 -6247.6
+ genre_TV.Movie 1 0.07 1303.5 -6247.5
+ original_isEnglish 1 0.01 1303.5 -6247.3
+ runtime 1 0.00 1303.5 -6247.3
+ num_lang 4 1.28 1302.3 -6246.0
- holiday_month 1 16.80 1320.3 -6189.8
- genre_Family 1 18.38 1321.9 -6184.0
- budget 1 444.14 1747.7 -4843.6
- vote_count 1 903.26 2206.8 -3723.7
Step: AIC=-6285.8
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector
Df Sum of Sq RSS AIC
+ genre_Science.Fiction 1 8.56 1284.6 -6315.7
+ topActor 1 7.86 1285.3 -6313.1
+ genre_Crime 1 5.96 1287.2 -6306.0
+ genre_Animation 1 4.43 1288.7 -6300.3
+ country_Germany 1 4.25 1288.9 -6299.6
+ genre_Romance 1 4.09 1289.0 -6299.0
+ genre_Thriller 1 3.70 1289.4 -6297.5
+ genre_Action 1 3.55 1289.6 -6297.0
+ popularity 1 3.41 1289.7 -6296.5
+ CastGenderAVG 1 3.01 1290.1 -6295.0
+ genre_Adventure 1 2.76 1290.3 -6294.1
+ genre_Western 1 2.48 1290.6 -6293.0
+ genre_Comedy 1 2.43 1290.7 -6292.8
+ country_France 1 2.32 1290.8 -6292.4
+ genre_Drama 1 2.21 1290.9 -6292.0
+ genre_Mystery 1 1.11 1292.0 -6287.9
+ country_Austr 1 1.10 1292.0 -6287.9
+ genre_Documentary 1 1.02 1292.1 -6287.6
+ genre_History 1 0.74 1292.4 -6286.5
+ genre_Music 1 0.74 1292.4 -6286.5
+ country_Canada 1 0.73 1292.4 -6286.5
+ genre_War 1 0.71 1292.4 -6286.4
<none> 1293.1 -6285.8
+ country_UK 1 0.47 1292.7 -6285.5
+ vote_average 1 0.27 1292.8 -6284.8
+ genre_Foreign 1 0.24 1292.9 -6284.7
+ genre_Horror 1 0.11 1293.0 -6284.2
+ runtime 1 0.11 1293.0 -6284.2
+ genre_TV.Movie 1 0.08 1293.0 -6284.1
+ country_USA 1 0.07 1293.0 -6284.1
+ genre_Fantasy 1 0.02 1293.1 -6283.9
+ original_isEnglish 1 0.00 1293.1 -6283.8
+ num_lang 4 1.25 1291.9 -6282.4
- topDirector 1 10.42 1303.5 -6249.3
- genre_Family 1 16.77 1309.9 -6225.9
- holiday_month 1 17.45 1310.6 -6223.4
- budget 1 440.53 1733.7 -4880.3
- vote_count 1 893.93 2187.1 -3764.9
Step: AIC=-6315.67
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction
Df Sum of Sq RSS AIC
+ topActor 1 9.34 1275.2 -6348.7
+ genre_Crime 1 7.94 1276.6 -6343.4
+ genre_Adventure 1 4.48 1280.1 -6330.4
+ country_Germany 1 4.45 1280.1 -6330.3
+ genre_Animation 1 4.25 1280.3 -6329.6
+ genre_Drama 1 4.14 1280.4 -6329.2
+ popularity 1 3.37 1281.2 -6326.3
+ genre_Romance 1 3.05 1281.5 -6325.1
+ genre_Western 1 2.88 1281.7 -6324.5
+ genre_Thriller 1 2.77 1281.8 -6324.0
+ CastGenderAVG 1 2.47 1282.1 -6322.9
+ country_France 1 2.42 1282.2 -6322.7
+ genre_Action 1 1.93 1282.6 -6320.9
+ genre_Comedy 1 1.68 1282.9 -6320.0
+ genre_History 1 1.15 1283.4 -6318.0
+ genre_Music 1 1.15 1283.4 -6318.0
+ genre_War 1 1.05 1283.5 -6317.6
+ genre_Mystery 1 1.03 1283.5 -6317.5
+ country_Austr 1 0.90 1283.7 -6317.0
+ genre_Documentary 1 0.83 1283.7 -6316.8
+ vote_average 1 0.61 1284.0 -6315.9
<none> 1284.6 -6315.7
+ country_UK 1 0.48 1284.1 -6315.5
+ country_Canada 1 0.47 1284.1 -6315.4
+ genre_Horror 1 0.40 1284.2 -6315.2
+ genre_Foreign 1 0.19 1284.4 -6314.4
+ country_USA 1 0.10 1284.5 -6314.1
+ genre_TV.Movie 1 0.09 1284.5 -6314.0
+ genre_Fantasy 1 0.06 1284.5 -6313.9
+ runtime 1 0.01 1284.5 -6313.7
+ original_isEnglish 1 0.00 1284.6 -6313.7
+ num_lang 4 1.29 1283.3 -6312.5
- genre_Science.Fiction 1 8.56 1293.1 -6285.8
- topDirector 1 10.76 1295.3 -6277.6
- genre_Family 1 16.05 1300.6 -6258.1
- holiday_month 1 17.29 1301.8 -6253.5
- budget 1 448.31 1732.9 -4880.4
- vote_count 1 902.13 2186.7 -3763.7
Step: AIC=-6348.72
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor
Df Sum of Sq RSS AIC
+ genre_Crime 1 6.53 1268.7 -6371.4
+ country_Germany 1 4.26 1271.0 -6362.8
+ genre_Animation 1 4.10 1271.1 -6362.2
+ popularity 1 3.89 1271.3 -6361.4
+ genre_Adventure 1 3.61 1271.6 -6360.3
+ genre_Romance 1 3.37 1271.8 -6359.4
+ genre_Western 1 2.81 1272.4 -6357.3
+ genre_Thriller 1 2.69 1272.5 -6356.9
+ genre_Action 1 2.65 1272.6 -6356.7
+ CastGenderAVG 1 2.58 1272.6 -6356.4
+ genre_Drama 1 2.38 1272.8 -6355.7
+ country_France 1 2.27 1273.0 -6355.3
+ genre_Comedy 1 1.59 1273.6 -6352.7
+ genre_History 1 1.07 1274.1 -6350.8
+ genre_Music 1 1.07 1274.1 -6350.8
+ genre_War 1 1.03 1274.2 -6350.6
+ genre_Mystery 1 0.89 1274.3 -6350.1
+ country_Austr 1 0.83 1274.4 -6349.9
+ genre_Documentary 1 0.59 1274.6 -6348.9
<none> 1275.2 -6348.7
+ country_Canada 1 0.52 1274.7 -6348.7
+ country_UK 1 0.31 1274.9 -6347.9
+ runtime 1 0.21 1275.0 -6347.5
+ vote_average 1 0.19 1275.0 -6347.4
+ original_isEnglish 1 0.13 1275.1 -6347.2
+ genre_TV.Movie 1 0.13 1275.1 -6347.2
+ genre_Foreign 1 0.10 1275.1 -6347.1
+ genre_Horror 1 0.08 1275.1 -6347.0
+ genre_Fantasy 1 0.02 1275.2 -6346.8
+ country_USA 1 0.00 1275.2 -6346.7
+ num_lang 4 1.16 1274.0 -6345.1
- topDirector 1 8.14 1283.4 -6320.2
- topActor 1 9.34 1284.6 -6315.7
- genre_Science.Fiction 1 10.04 1285.3 -6313.1
- genre_Family 1 13.65 1288.9 -6299.6
- holiday_month 1 17.98 1293.2 -6283.5
- budget 1 456.09 1731.3 -4882.8
- vote_count 1 911.47 2186.7 -3761.7
Step: AIC=-6371.38
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime
Df Sum of Sq RSS AIC
+ country_Germany 1 3.98 1264.7 -6384.5
+ popularity 1 3.94 1264.7 -6384.3
+ genre_Animation 1 3.78 1264.9 -6383.7
+ genre_Adventure 1 3.12 1265.6 -6381.2
+ genre_Western 1 3.00 1265.7 -6380.7
+ genre_Drama 1 2.36 1266.3 -6378.3
+ genre_Romance 1 2.14 1266.5 -6377.5
+ country_France 1 1.97 1266.7 -6376.8
+ CastGenderAVG 1 1.57 1267.1 -6375.3
+ genre_War 1 1.46 1267.2 -6374.9
+ genre_History 1 1.45 1267.2 -6374.9
+ genre_Music 1 1.45 1267.2 -6374.9
+ genre_Action 1 1.41 1267.3 -6374.7
+ genre_Comedy 1 1.14 1267.5 -6373.7
+ country_Austr 1 0.90 1267.8 -6372.8
+ genre_Thriller 1 0.82 1267.9 -6372.5
<none> 1268.7 -6371.4
+ country_Canada 1 0.49 1268.2 -6371.3
+ genre_Mystery 1 0.43 1268.2 -6371.0
+ genre_Documentary 1 0.37 1268.3 -6370.8
+ country_UK 1 0.25 1268.4 -6370.3
+ runtime 1 0.24 1268.4 -6370.3
+ original_isEnglish 1 0.13 1268.5 -6369.9
+ vote_average 1 0.12 1268.6 -6369.8
+ genre_TV.Movie 1 0.12 1268.6 -6369.8
+ genre_Foreign 1 0.09 1268.6 -6369.7
+ genre_Horror 1 0.01 1268.7 -6369.4
+ genre_Fantasy 1 0.00 1268.7 -6369.4
+ country_USA 1 0.00 1268.7 -6369.4
+ num_lang 4 1.09 1267.6 -6367.5
- genre_Crime 1 6.53 1275.2 -6348.7
- topDirector 1 7.83 1276.5 -6343.8
- topActor 1 7.94 1276.6 -6343.4
- genre_Family 1 11.34 1280.0 -6330.7
- genre_Science.Fiction 1 11.83 1280.5 -6328.8
- holiday_month 1 16.94 1285.6 -6309.7
- budget 1 455.35 1724.0 -4901.0
- vote_count 1 914.62 2183.3 -3767.1
Step: AIC=-6384.46
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany
Df Sum of Sq RSS AIC
+ popularity 1 3.93 1260.8 -6397.4
+ genre_Animation 1 3.68 1261.0 -6396.5
+ genre_Western 1 3.11 1261.6 -6394.3
+ genre_Adventure 1 3.01 1261.7 -6393.9
+ genre_Drama 1 2.15 1262.5 -6390.6
+ genre_Romance 1 2.08 1262.6 -6390.4
+ CastGenderAVG 1 1.45 1263.3 -6388.0
+ country_France 1 1.43 1263.3 -6387.9
+ genre_Action 1 1.32 1263.4 -6387.5
+ genre_History 1 1.23 1263.5 -6387.1
+ genre_Music 1 1.23 1263.5 -6387.1
+ genre_War 1 1.20 1263.5 -6387.0
+ genre_Comedy 1 1.00 1263.7 -6386.3
+ country_Austr 1 0.99 1263.7 -6386.2
+ genre_Thriller 1 0.68 1264.0 -6385.0
<none> 1264.7 -6384.5
+ country_Canada 1 0.44 1264.3 -6384.1
+ runtime 1 0.38 1264.3 -6383.9
+ genre_Mystery 1 0.38 1264.3 -6383.9
+ genre_Documentary 1 0.33 1264.4 -6383.7
+ genre_TV.Movie 1 0.13 1264.6 -6383.0
+ genre_Foreign 1 0.07 1264.6 -6382.7
+ vote_average 1 0.07 1264.6 -6382.7
+ country_UK 1 0.05 1264.7 -6382.6
+ original_isEnglish 1 0.03 1264.7 -6382.6
+ genre_Horror 1 0.01 1264.7 -6382.5
+ country_USA 1 0.01 1264.7 -6382.5
+ genre_Fantasy 1 0.00 1264.7 -6382.5
+ num_lang 4 1.13 1263.6 -6380.7
- country_Germany 1 3.98 1268.7 -6371.4
- genre_Crime 1 6.25 1271.0 -6362.8
- topActor 1 7.80 1272.5 -6356.9
- topDirector 1 8.04 1272.7 -6356.0
- genre_Family 1 10.76 1275.5 -6345.8
- genre_Science.Fiction 1 11.99 1276.7 -6341.2
- holiday_month 1 16.21 1280.9 -6325.3
- budget 1 459.13 1723.8 -4899.5
- vote_count 1 910.72 2175.4 -3782.5
Step: AIC=-6397.41
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity
Df Sum of Sq RSS AIC
+ genre_Animation 1 3.30 1257.5 -6408.0
+ genre_Western 1 3.04 1257.7 -6407.0
+ genre_Adventure 1 2.73 1258.0 -6405.8
+ genre_Romance 1 2.17 1258.6 -6403.7
+ genre_Drama 1 2.04 1258.7 -6403.2
+ country_France 1 1.49 1259.3 -6401.1
+ CastGenderAVG 1 1.44 1259.3 -6400.9
+ genre_Action 1 1.32 1259.5 -6400.4
+ genre_War 1 1.23 1259.5 -6400.1
+ genre_History 1 1.22 1259.5 -6400.1
+ genre_Music 1 1.22 1259.5 -6400.1
+ country_Austr 1 1.04 1259.7 -6399.4
+ genre_Comedy 1 1.00 1259.8 -6399.2
+ genre_Thriller 1 0.84 1259.9 -6398.6
<none> 1260.8 -6397.4
+ country_Canada 1 0.49 1260.3 -6397.3
+ genre_Mystery 1 0.42 1260.4 -6397.0
+ genre_Documentary 1 0.40 1260.4 -6396.9
+ runtime 1 0.36 1260.4 -6396.8
+ vote_average 1 0.14 1260.6 -6395.9
+ genre_TV.Movie 1 0.12 1260.7 -6395.8
+ genre_Foreign 1 0.10 1260.7 -6395.8
+ country_UK 1 0.07 1260.7 -6395.7
+ country_USA 1 0.03 1260.7 -6395.5
+ original_isEnglish 1 0.03 1260.8 -6395.5
+ genre_Fantasy 1 0.00 1260.8 -6395.4
+ genre_Horror 1 0.00 1260.8 -6395.4
+ num_lang 4 1.23 1259.5 -6394.1
- popularity 1 3.93 1264.7 -6384.5
- country_Germany 1 3.97 1264.7 -6384.3
- genre_Crime 1 6.29 1267.1 -6375.5
- topDirector 1 7.81 1268.6 -6369.8
- topActor 1 8.28 1269.1 -6368.0
- genre_Family 1 10.45 1271.2 -6359.8
- genre_Science.Fiction 1 11.99 1272.8 -6354.0
- holiday_month 1 15.94 1276.7 -6339.1
- vote_count 1 441.96 1702.7 -4956.6
- budget 1 451.31 1712.1 -4930.4
Step: AIC=-6408.01
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation
Df Sum of Sq RSS AIC
+ genre_Western 1 3.04 1254.4 -6417.6
+ genre_Adventure 1 2.49 1255.0 -6415.5
+ genre_Romance 1 2.46 1255.0 -6415.4
+ country_France 1 1.62 1255.8 -6412.2
+ genre_Drama 1 1.61 1255.9 -6412.2
+ CastGenderAVG 1 1.60 1255.9 -6412.1
+ genre_War 1 1.18 1256.3 -6410.5
+ genre_History 1 1.15 1256.3 -6410.4
+ genre_Music 1 1.15 1256.3 -6410.4
+ genre_Action 1 1.07 1256.4 -6410.1
+ country_Austr 1 0.99 1256.5 -6409.8
+ genre_Comedy 1 0.98 1256.5 -6409.7
+ runtime 1 0.82 1256.7 -6409.2
+ genre_Thriller 1 0.73 1256.7 -6408.8
<none> 1257.5 -6408.0
+ country_Canada 1 0.46 1257.0 -6407.8
+ genre_Documentary 1 0.43 1257.0 -6407.6
+ genre_Mystery 1 0.37 1257.1 -6407.4
+ vote_average 1 0.23 1257.2 -6406.9
+ original_isEnglish 1 0.12 1257.3 -6406.4
+ genre_TV.Movie 1 0.12 1257.4 -6406.4
+ genre_Foreign 1 0.10 1257.4 -6406.4
+ country_UK 1 0.03 1257.4 -6406.1
+ genre_Horror 1 0.00 1257.5 -6406.0
+ country_USA 1 0.00 1257.5 -6406.0
+ genre_Fantasy 1 0.00 1257.5 -6406.0
+ num_lang 4 1.18 1256.3 -6404.5
- genre_Animation 1 3.30 1260.8 -6397.4
- genre_Family 1 3.53 1261.0 -6396.5
- popularity 1 3.55 1261.0 -6396.5
- country_Germany 1 3.88 1261.3 -6395.2
- genre_Crime 1 6.00 1263.5 -6387.2
- topDirector 1 7.67 1265.1 -6380.8
- topActor 1 8.16 1265.6 -6378.9
- genre_Science.Fiction 1 11.74 1269.2 -6365.4
- holiday_month 1 15.72 1273.2 -6350.3
- budget 1 439.47 1696.9 -4971.0
- vote_count 1 443.64 1701.1 -4959.2
Step: AIC=-6417.61
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western
Df Sum of Sq RSS AIC
+ genre_Adventure 1 2.81 1251.6 -6426.4
+ genre_Romance 1 2.33 1252.1 -6424.5
+ genre_Drama 1 1.76 1252.7 -6422.4
+ country_France 1 1.62 1252.8 -6421.8
+ genre_War 1 1.19 1253.2 -6420.2
+ CastGenderAVG 1 1.11 1253.3 -6419.8
+ genre_History 1 1.04 1253.4 -6419.6
+ genre_Music 1 1.04 1253.4 -6419.6
+ runtime 1 1.03 1253.4 -6419.5
+ country_Austr 1 0.94 1253.5 -6419.2
+ genre_Thriller 1 0.89 1253.5 -6419.0
+ genre_Comedy 1 0.84 1253.6 -6418.8
+ genre_Action 1 0.83 1253.6 -6418.8
<none> 1254.4 -6417.6
+ genre_Mystery 1 0.42 1254.0 -6417.2
+ country_Canada 1 0.41 1254.0 -6417.2
+ genre_Documentary 1 0.37 1254.1 -6417.0
+ vote_average 1 0.21 1254.2 -6416.4
+ genre_TV.Movie 1 0.12 1254.3 -6416.1
+ original_isEnglish 1 0.11 1254.3 -6416.0
+ genre_Foreign 1 0.08 1254.3 -6415.9
+ country_UK 1 0.04 1254.4 -6415.8
+ genre_Fantasy 1 0.00 1254.4 -6415.6
+ country_USA 1 0.00 1254.4 -6415.6
+ genre_Horror 1 0.00 1254.4 -6415.6
+ num_lang 4 1.15 1253.3 -6414.0
- genre_Western 1 3.04 1257.5 -6408.0
- genre_Animation 1 3.30 1257.7 -6407.0
- genre_Family 1 3.34 1257.8 -6406.9
- popularity 1 3.48 1257.9 -6406.3
- country_Germany 1 4.00 1258.4 -6404.3
- genre_Crime 1 6.17 1260.6 -6396.1
- topDirector 1 7.85 1262.3 -6389.7
- topActor 1 8.07 1262.5 -6388.8
- genre_Science.Fiction 1 12.24 1266.7 -6373.0
- holiday_month 1 16.15 1270.6 -6358.2
- budget 1 439.74 1694.2 -4976.9
- vote_count 1 444.17 1698.6 -4964.3
Step: AIC=-6426.39
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure
Df Sum of Sq RSS AIC
+ genre_Romance 1 2.67 1249.0 -6434.6
+ genre_Action 1 2.20 1249.4 -6432.8
+ CastGenderAVG 1 1.69 1249.9 -6430.9
+ country_France 1 1.63 1250.0 -6430.7
+ genre_Drama 1 1.39 1250.2 -6429.7
+ genre_War 1 1.30 1250.3 -6429.4
+ genre_Comedy 1 1.07 1250.5 -6428.5
+ genre_History 1 1.05 1250.6 -6428.4
+ genre_Music 1 1.05 1250.6 -6428.4
+ country_Austr 1 0.98 1250.6 -6428.2
+ genre_Thriller 1 0.94 1250.7 -6428.0
+ runtime 1 0.93 1250.7 -6427.9
<none> 1251.6 -6426.4
+ genre_Documentary 1 0.41 1251.2 -6426.0
+ country_Canada 1 0.40 1251.2 -6425.9
+ genre_Mystery 1 0.35 1251.3 -6425.8
+ vote_average 1 0.21 1251.4 -6425.2
+ original_isEnglish 1 0.14 1251.5 -6424.9
+ country_UK 1 0.12 1251.5 -6424.8
+ genre_TV.Movie 1 0.10 1251.5 -6424.8
+ genre_Foreign 1 0.08 1251.5 -6424.7
+ genre_Fantasy 1 0.05 1251.6 -6424.6
+ genre_Horror 1 0.02 1251.6 -6424.5
+ country_USA 1 0.00 1251.6 -6424.4
+ num_lang 4 1.13 1250.5 -6422.7
- genre_Family 1 2.58 1254.2 -6418.5
- genre_Adventure 1 2.81 1254.4 -6417.6
- genre_Animation 1 3.05 1254.7 -6416.7
- popularity 1 3.22 1254.8 -6416.1
- genre_Western 1 3.36 1255.0 -6415.5
- country_Germany 1 3.90 1255.5 -6413.4
- genre_Crime 1 5.74 1257.4 -6406.4
- topActor 1 7.38 1259.0 -6400.2
- topDirector 1 7.64 1259.3 -6399.2
- genre_Science.Fiction 1 13.67 1265.3 -6376.3
- holiday_month 1 16.00 1267.6 -6367.4
- budget 1 387.46 1639.1 -5133.6
- vote_count 1 443.83 1695.5 -4971.2
Step: AIC=-6434.64
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance
Df Sum of Sq RSS AIC
+ genre_Drama 1 2.01 1246.9 -6440.4
+ country_France 1 1.67 1247.3 -6439.1
+ genre_Action 1 1.64 1247.3 -6439.0
+ genre_War 1 1.23 1247.7 -6437.4
+ CastGenderAVG 1 1.06 1247.9 -6436.7
+ country_Austr 1 0.93 1248.0 -6436.2
+ genre_History 1 0.92 1248.0 -6436.2
+ genre_Music 1 0.92 1248.0 -6436.2
+ runtime 1 0.73 1248.2 -6435.4
+ genre_Documentary 1 0.66 1248.3 -6435.2
+ genre_Comedy 1 0.59 1248.4 -6434.9
<none> 1249.0 -6434.6
+ genre_Thriller 1 0.47 1248.5 -6434.4
+ vote_average 1 0.32 1248.6 -6433.9
+ country_Canada 1 0.32 1248.6 -6433.9
+ genre_Mystery 1 0.23 1248.7 -6433.5
+ genre_Horror 1 0.19 1248.8 -6433.4
+ country_UK 1 0.14 1248.8 -6433.2
+ original_isEnglish 1 0.14 1248.8 -6433.2
+ genre_TV.Movie 1 0.13 1248.8 -6433.1
+ genre_Foreign 1 0.08 1248.9 -6432.9
+ genre_Fantasy 1 0.06 1248.9 -6432.9
+ country_USA 1 0.00 1249.0 -6432.6
+ num_lang 4 1.19 1247.8 -6431.2
- genre_Romance 1 2.67 1251.6 -6426.4
- genre_Family 1 2.76 1251.7 -6426.1
- genre_Adventure 1 3.16 1252.1 -6424.5
- genre_Western 1 3.23 1252.2 -6424.3
- popularity 1 3.27 1252.2 -6424.1
- genre_Animation 1 3.34 1252.3 -6423.8
- country_Germany 1 3.82 1252.8 -6422.0
- genre_Crime 1 4.45 1253.4 -6419.6
- topActor 1 7.72 1256.7 -6407.1
- topDirector 1 7.73 1256.7 -6407.0
- genre_Science.Fiction 1 12.29 1261.2 -6389.6
- holiday_month 1 15.73 1264.7 -6376.5
- budget 1 388.69 1637.6 -5135.8
- vote_count 1 444.81 1693.8 -4974.0
Step: AIC=-6440.39
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama
Df Sum of Sq RSS AIC
+ genre_Action 1 2.10 1244.8 -6446.5
+ runtime 1 1.76 1245.2 -6445.2
+ country_France 1 1.56 1245.4 -6444.4
+ CastGenderAVG 1 1.06 1245.9 -6442.5
+ country_Austr 1 0.98 1246.0 -6442.2
+ genre_War 1 0.90 1246.0 -6441.9
+ genre_Thriller 1 0.55 1246.4 -6440.5
+ genre_History 1 0.55 1246.4 -6440.5
+ genre_Music 1 0.55 1246.4 -6440.5
<none> 1246.9 -6440.4
+ genre_Documentary 1 0.37 1246.6 -6439.8
+ country_Canada 1 0.33 1246.6 -6439.6
+ genre_Mystery 1 0.21 1246.7 -6439.2
+ genre_Comedy 1 0.15 1246.8 -6439.0
+ genre_Foreign 1 0.12 1246.8 -6438.9
+ genre_TV.Movie 1 0.12 1246.8 -6438.8
+ genre_Fantasy 1 0.11 1246.8 -6438.8
+ country_UK 1 0.10 1246.8 -6438.8
+ vote_average 1 0.05 1246.9 -6438.6
+ original_isEnglish 1 0.05 1246.9 -6438.6
+ country_USA 1 0.02 1246.9 -6438.5
+ genre_Horror 1 0.01 1246.9 -6438.4
+ num_lang 4 1.22 1245.7 -6437.1
- genre_Drama 1 2.01 1249.0 -6434.6
- genre_Family 1 2.53 1249.5 -6432.6
- genre_Adventure 1 2.72 1249.7 -6431.9
- genre_Animation 1 2.91 1249.8 -6431.2
- popularity 1 3.22 1250.2 -6430.0
- genre_Romance 1 3.30 1250.2 -6429.7
- genre_Western 1 3.35 1250.3 -6429.5
- country_Germany 1 3.63 1250.6 -6428.4
- genre_Crime 1 4.36 1251.3 -6425.6
- topActor 1 6.34 1253.3 -6418.0
- topDirector 1 7.11 1254.0 -6415.1
- genre_Science.Fiction 1 13.47 1260.4 -6390.8
- holiday_month 1 16.34 1263.3 -6379.9
- budget 1 377.70 1624.6 -5172.1
- vote_count 1 443.79 1690.7 -4980.6
Step: AIC=-6446.47
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action
Df Sum of Sq RSS AIC
+ runtime 1 1.81 1243.0 -6451.5
+ country_France 1 1.49 1243.3 -6450.2
+ country_Austr 1 0.92 1243.9 -6448.0
+ genre_War 1 0.71 1244.1 -6447.2
+ CastGenderAVG 1 0.65 1244.2 -6447.0
<none> 1244.8 -6446.5
+ genre_History 1 0.44 1244.4 -6446.2
+ genre_Music 1 0.44 1244.4 -6446.2
+ genre_Mystery 1 0.37 1244.5 -6445.9
+ country_Canada 1 0.32 1244.5 -6445.7
+ genre_Documentary 1 0.27 1244.6 -6445.5
+ genre_Thriller 1 0.26 1244.6 -6445.5
+ genre_Foreign 1 0.13 1244.7 -6445.0
+ genre_Fantasy 1 0.12 1244.7 -6444.9
+ genre_TV.Movie 1 0.10 1244.7 -6444.9
+ country_UK 1 0.10 1244.7 -6444.9
+ vote_average 1 0.08 1244.8 -6444.8
+ genre_Comedy 1 0.07 1244.8 -6444.7
+ original_isEnglish 1 0.03 1244.8 -6444.6
+ country_USA 1 0.03 1244.8 -6444.6
+ genre_Horror 1 0.00 1244.8 -6444.5
+ num_lang 4 1.15 1243.7 -6442.9
- genre_Family 1 1.84 1246.7 -6441.4
- genre_Action 1 2.10 1246.9 -6440.4
- genre_Animation 1 2.43 1247.3 -6439.1
- genre_Drama 1 2.47 1247.3 -6439.0
- genre_Romance 1 2.69 1247.5 -6438.1
- genre_Western 1 3.06 1247.9 -6436.7
- popularity 1 3.16 1248.0 -6436.3
- genre_Crime 1 3.18 1248.0 -6436.2
- country_Germany 1 3.48 1248.3 -6435.1
- genre_Adventure 1 3.99 1248.8 -6433.1
- topActor 1 6.73 1251.6 -6422.6
- topDirector 1 7.50 1252.3 -6419.6
- genre_Science.Fiction 1 11.89 1256.7 -6402.8
- holiday_month 1 16.28 1261.1 -6386.1
- budget 1 375.42 1620.3 -5183.0
- vote_count 1 444.13 1689.0 -4983.6
Step: AIC=-6451.46
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime
Df Sum of Sq RSS AIC
+ country_France 1 1.59 1241.4 -6455.6
+ genre_War 1 1.14 1241.9 -6453.9
+ genre_History 1 1.02 1242.0 -6453.4
+ genre_Music 1 1.02 1242.0 -6453.4
+ CastGenderAVG 1 0.97 1242.1 -6453.2
+ country_Austr 1 0.88 1242.1 -6452.9
<none> 1243.0 -6451.5
+ vote_average 1 0.48 1242.5 -6451.3
+ genre_Mystery 1 0.40 1242.6 -6451.0
+ genre_Documentary 1 0.28 1242.8 -6450.6
+ country_Canada 1 0.27 1242.8 -6450.5
+ genre_Thriller 1 0.23 1242.8 -6450.4
+ genre_Comedy 1 0.19 1242.8 -6450.2
+ country_UK 1 0.15 1242.9 -6450.0
+ genre_Fantasy 1 0.11 1242.9 -6449.9
+ original_isEnglish 1 0.11 1242.9 -6449.9
+ genre_Foreign 1 0.10 1242.9 -6449.9
+ genre_TV.Movie 1 0.08 1243.0 -6449.8
+ country_USA 1 0.03 1243.0 -6449.6
+ genre_Horror 1 0.00 1243.0 -6449.5
+ num_lang 4 1.43 1241.6 -6449.0
- runtime 1 1.81 1244.8 -6446.5
- genre_Family 1 2.09 1245.1 -6445.4
- genre_Action 1 2.15 1245.2 -6445.2
- genre_Romance 1 2.56 1245.6 -6443.6
- genre_Animation 1 3.01 1246.0 -6441.8
- popularity 1 3.05 1246.1 -6441.7
- genre_Crime 1 3.23 1246.3 -6441.0
- genre_Western 1 3.38 1246.4 -6440.4
- genre_Drama 1 3.59 1246.6 -6439.6
- genre_Adventure 1 3.68 1246.7 -6439.3
- country_Germany 1 3.76 1246.8 -6439.0
- topActor 1 7.15 1250.2 -6425.9
- topDirector 1 8.00 1251.0 -6422.7
- genre_Science.Fiction 1 11.72 1254.8 -6408.4
- holiday_month 1 15.09 1258.1 -6395.5
- budget 1 354.07 1597.1 -5250.1
- vote_count 1 437.58 1680.6 -5005.5
Step: AIC=-6455.6
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime + country_France
Df Sum of Sq RSS AIC
+ genre_War 1 1.06 1240.4 -6457.7
+ CastGenderAVG 1 1.00 1240.4 -6457.5
+ genre_History 1 0.97 1240.5 -6457.3
+ genre_Music 1 0.97 1240.5 -6457.3
+ country_Austr 1 0.91 1240.5 -6457.1
<none> 1241.4 -6455.6
+ vote_average 1 0.42 1241.0 -6455.2
+ genre_Mystery 1 0.38 1241.1 -6455.1
+ genre_Documentary 1 0.30 1241.1 -6454.8
+ country_USA 1 0.24 1241.2 -6454.5
+ genre_Thriller 1 0.21 1241.2 -6454.4
+ country_Canada 1 0.20 1241.2 -6454.4
+ genre_Comedy 1 0.13 1241.3 -6454.1
+ genre_Fantasy 1 0.10 1241.3 -6454.0
+ genre_Foreign 1 0.09 1241.3 -6454.0
+ genre_TV.Movie 1 0.09 1241.3 -6453.9
+ country_UK 1 0.03 1241.4 -6453.7
+ genre_Horror 1 0.00 1241.4 -6453.6
+ original_isEnglish 1 0.00 1241.4 -6453.6
+ num_lang 4 1.34 1240.1 -6452.8
- country_France 1 1.59 1243.0 -6451.5
- runtime 1 1.91 1243.3 -6450.2
- genre_Family 1 2.01 1243.5 -6449.8
- genre_Action 1 2.08 1243.5 -6449.6
- genre_Romance 1 2.59 1244.0 -6447.6
- genre_Crime 1 3.07 1244.5 -6445.8
- popularity 1 3.09 1244.5 -6445.7
- genre_Animation 1 3.17 1244.6 -6445.4
- country_Germany 1 3.23 1244.7 -6445.1
- genre_Western 1 3.39 1244.8 -6444.5
- genre_Drama 1 3.47 1244.9 -6444.2
- genre_Adventure 1 3.67 1245.1 -6443.4
- topActor 1 7.12 1248.6 -6430.1
- topDirector 1 8.11 1249.5 -6426.3
- genre_Science.Fiction 1 11.71 1253.2 -6412.5
- holiday_month 1 15.18 1256.6 -6399.3
- budget 1 351.43 1592.9 -5260.9
- vote_count 1 436.96 1678.4 -5009.8
Step: AIC=-6457.69
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime + country_France + genre_War
Df Sum of Sq RSS AIC
+ country_Austr 1 0.91 1239.5 -6459.2
+ CastGenderAVG 1 0.73 1239.7 -6458.5
+ genre_History 1 0.54 1239.8 -6457.8
+ genre_Music 1 0.54 1239.8 -6457.8
<none> 1240.4 -6457.7
+ genre_Mystery 1 0.42 1240.0 -6457.3
+ vote_average 1 0.38 1240.0 -6457.2
+ genre_Documentary 1 0.29 1240.1 -6456.8
+ country_USA 1 0.27 1240.1 -6456.7
+ genre_Thriller 1 0.26 1240.1 -6456.7
+ country_Canada 1 0.20 1240.2 -6456.5
+ genre_Fantasy 1 0.13 1240.3 -6456.2
+ genre_Comedy 1 0.10 1240.3 -6456.1
+ genre_TV.Movie 1 0.09 1240.3 -6456.0
+ genre_Foreign 1 0.08 1240.3 -6456.0
+ country_UK 1 0.02 1240.4 -6455.8
+ genre_Horror 1 0.00 1240.4 -6455.7
+ original_isEnglish 1 0.00 1240.4 -6455.7
- genre_War 1 1.06 1241.4 -6455.6
+ num_lang 4 1.43 1239.0 -6455.2
- country_France 1 1.51 1241.9 -6453.9
- genre_Action 1 1.86 1242.2 -6452.5
- genre_Family 1 1.91 1242.3 -6452.3
- runtime 1 2.32 1242.7 -6450.7
- genre_Romance 1 2.48 1242.9 -6450.1
- country_Germany 1 3.07 1243.5 -6447.8
- popularity 1 3.11 1243.5 -6447.7
- genre_Drama 1 3.18 1243.6 -6447.4
- genre_Animation 1 3.26 1243.6 -6447.1
- genre_Crime 1 3.39 1243.8 -6446.6
- genre_Western 1 3.45 1243.8 -6446.3
- genre_Adventure 1 3.70 1244.1 -6445.4
- topActor 1 7.19 1247.6 -6431.9
- topDirector 1 8.18 1248.6 -6428.1
- genre_Science.Fiction 1 12.13 1252.5 -6413.0
- holiday_month 1 15.31 1255.7 -6400.8
- budget 1 351.22 1591.6 -5262.7
- vote_count 1 436.06 1676.4 -5013.4
Step: AIC=-6459.2
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime + country_France + genre_War + country_Austr
Df Sum of Sq RSS AIC
+ CastGenderAVG 1 0.73 1238.8 -6460.0
+ genre_History 1 0.53 1238.9 -6459.3
+ genre_Music 1 0.53 1238.9 -6459.3
<none> 1239.5 -6459.2
+ genre_Mystery 1 0.40 1239.1 -6458.8
+ vote_average 1 0.37 1239.1 -6458.6
+ country_USA 1 0.35 1239.1 -6458.6
+ genre_Documentary 1 0.28 1239.2 -6458.3
+ genre_Thriller 1 0.24 1239.2 -6458.1
+ country_Canada 1 0.20 1239.3 -6458.0
+ genre_Fantasy 1 0.13 1239.3 -6457.7
- country_Austr 1 0.91 1240.4 -6457.7
+ genre_Foreign 1 0.10 1239.4 -6457.6
+ genre_TV.Movie 1 0.09 1239.4 -6457.6
+ genre_Comedy 1 0.08 1239.4 -6457.5
+ country_UK 1 0.02 1239.5 -6457.3
+ genre_Horror 1 0.00 1239.5 -6457.2
+ original_isEnglish 1 0.00 1239.5 -6457.2
- genre_War 1 1.05 1240.5 -6457.1
+ num_lang 4 1.39 1238.1 -6456.6
- country_France 1 1.53 1241.0 -6455.3
- genre_Action 1 1.81 1241.3 -6454.2
- genre_Family 1 1.94 1241.4 -6453.7
- runtime 1 2.28 1241.8 -6452.4
- genre_Romance 1 2.45 1241.9 -6451.7
- country_Germany 1 3.14 1242.6 -6449.1
- popularity 1 3.15 1242.6 -6449.0
- genre_Animation 1 3.21 1242.7 -6448.8
- genre_Drama 1 3.21 1242.7 -6448.8
- genre_Western 1 3.41 1242.9 -6448.0
- genre_Crime 1 3.46 1242.9 -6447.8
- genre_Adventure 1 3.71 1243.2 -6446.8
- topActor 1 7.09 1246.6 -6433.8
- topDirector 1 8.20 1247.7 -6429.6
- genre_Science.Fiction 1 11.96 1251.4 -6415.1
- holiday_month 1 15.21 1254.7 -6402.7
- budget 1 352.01 1591.5 -5261.1
- vote_count 1 434.59 1674.1 -5018.2
Step: AIC=-6460.02
revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime + country_France + genre_War + country_Austr + CastGenderAVG
Df Sum of Sq RSS AIC
<none> 1238.8 -6460.0
+ genre_History 1 0.45 1238.3 -6459.8
+ genre_Music 1 0.45 1238.3 -6459.8
+ genre_Mystery 1 0.42 1238.3 -6459.6
+ genre_Documentary 1 0.37 1238.4 -6459.5
+ country_USA 1 0.37 1238.4 -6459.4
- CastGenderAVG 1 0.73 1239.5 -6459.2
+ vote_average 1 0.27 1238.5 -6459.1
- genre_War 1 0.78 1239.5 -6459.0
+ genre_Thriller 1 0.22 1238.5 -6458.9
+ country_Canada 1 0.22 1238.5 -6458.9
- country_Austr 1 0.91 1239.7 -6458.5
+ genre_Fantasy 1 0.12 1238.6 -6458.5
+ genre_TV.Movie 1 0.10 1238.7 -6458.4
+ genre_Foreign 1 0.09 1238.7 -6458.4
+ genre_Comedy 1 0.06 1238.7 -6458.3
+ country_UK 1 0.01 1238.7 -6458.1
+ original_isEnglish 1 0.00 1238.8 -6458.0
+ genre_Horror 1 0.00 1238.8 -6458.0
+ num_lang 4 1.44 1237.3 -6457.6
- genre_Action 1 1.44 1240.2 -6456.5
- country_France 1 1.57 1240.3 -6456.0
- genre_Family 1 1.95 1240.7 -6454.5
- genre_Romance 1 1.98 1240.7 -6454.3
- runtime 1 2.53 1241.3 -6452.2
- genre_Western 1 3.04 1241.8 -6450.2
- genre_Crime 1 3.07 1241.8 -6450.1
- country_Germany 1 3.10 1241.8 -6450.0
- popularity 1 3.11 1241.9 -6450.0
- genre_Drama 1 3.30 1242.0 -6449.2
- genre_Animation 1 3.36 1242.1 -6449.0
- genre_Adventure 1 3.94 1242.7 -6446.8
- topActor 1 7.12 1245.9 -6434.5
- topDirector 1 7.98 1246.7 -6431.2
- genre_Science.Fiction 1 11.76 1250.5 -6416.6
- holiday_month 1 15.30 1254.0 -6403.1
- budget 1 351.97 1590.7 -5261.4
- vote_count 1 434.23 1673.0 -5019.3
Call:
lm(formula = revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime + country_France + genre_War + country_Austr + CastGenderAVG,
data = tmdb1)
Coefficients:
(Intercept) vote_count budget genre_Family1 holiday_month1
0.05092 0.53105 0.38772 0.08003 0.12465
topDirector1 genre_Science.Fiction1 topActor1 genre_Crime1 country_Germany1
-0.15669 -0.16967 -0.09180 -0.07667 -0.10273
popularity genre_Animation1 genre_Western1 genre_Adventure1 genre_Romance1
0.04088 0.15026 -0.19713 0.09240 0.05540
genre_Drama1 genre_Action1 runtime country_France1 genre_War1
-0.06040 -0.04868 0.02730 -0.07472 -0.07833
country_Austr1 CastGenderAVG
-0.09208 -0.01349 summary(full_model)
Call:
lm(formula = revenue ~ ., data = tmdb1)
Residuals:
Min 1Q Median 3Q Max
-3.8025 -0.1693 0.0017 0.1322 9.7054
Coefficients: (1 not defined because of singularities)
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.097735 0.069251 1.411 0.158218
topDirector1 -0.157837 0.028409 -5.556 2.91e-08 ***
topActor1 -0.088891 0.017824 -4.987 6.34e-07 ***
CastGenderAVG -0.012994 0.008203 -1.584 0.113245
budget 0.389801 0.010915 35.712 < 2e-16 ***
popularity 0.044147 0.011870 3.719 0.000202 ***
runtime 0.034621 0.009499 3.645 0.000271 ***
vote_average -0.007514 0.008898 -0.844 0.398442
vote_count 0.530880 0.013282 39.969 < 2e-16 ***
holiday_month1 0.123663 0.016298 7.588 3.89e-14 ***
num_lang1 -0.051374 0.061032 -0.842 0.399971
num_lang2 -0.030458 0.063359 -0.481 0.630742
num_lang3 -0.079123 0.067955 -1.164 0.244343
num_lang4 -0.003689 0.071847 -0.051 0.959051
original_isEnglish1 0.025913 0.038366 0.675 0.499442
country_USA1 -0.032793 0.025847 -1.269 0.204603
country_UK1 -0.011485 0.024131 -0.476 0.634117
country_Germany1 -0.100186 0.030246 -3.312 0.000932 ***
country_France1 -0.073767 0.032297 -2.284 0.022415 *
country_Canada1 -0.033457 0.033374 -1.002 0.316161
country_Austr1 -0.094867 0.049595 -1.913 0.055830 .
genre_Action1 -0.043894 0.021654 -2.027 0.042714 *
genre_Adventure1 0.093637 0.024105 3.885 0.000104 ***
genre_Fantasy1 -0.019503 0.028095 -0.694 0.487603
genre_Science.Fiction1 -0.164991 0.025564 -6.454 1.20e-10 ***
genre_Crime1 -0.068867 0.023378 -2.946 0.003236 **
genre_Drama1 -0.050728 0.018944 -2.678 0.007436 **
genre_Thriller1 -0.013727 0.020770 -0.661 0.508686
genre_Animation1 0.155027 0.042564 3.642 0.000273 ***
genre_Family1 0.082812 0.030178 2.744 0.006090 **
genre_Western1 -0.193558 0.057979 -3.338 0.000849 ***
genre_Comedy1 0.009134 0.018760 0.487 0.626351
genre_Romance1 0.053442 0.020785 2.571 0.010167 *
genre_Horror1 0.019980 0.027710 0.721 0.470921
genre_Mystery1 -0.029991 0.030161 -0.994 0.320093
genre_History1 -0.064536 0.041295 -1.563 0.118164
genre_War1 -0.064576 0.047029 -1.373 0.169781
genre_Music1 NA NA NA NA
genre_Documentary1 0.064657 0.053128 1.217 0.223659
genre_Foreign1 0.032684 0.089623 0.365 0.715361
genre_TV.Movie1 -0.112887 0.180786 -0.624 0.532380
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.5092 on 4761 degrees of freedom
Multiple R-squared: 0.7428, Adjusted R-squared: 0.7407
F-statistic: 352.5 on 39 and 4761 DF, p-value: < 2.2e-16dflm1 <- lm(revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime, data = tmdb1)
summary(dflm1)
Call:
lm(formula = revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector + genre_Science.Fiction + topActor + genre_Crime +
country_Germany + popularity + genre_Animation + genre_Western +
genre_Adventure + genre_Romance + genre_Drama + genre_Action +
runtime, data = tmdb1)
Residuals:
Min 1Q Median 3Q Max
-3.7865 -0.1655 0.0053 0.1315 9.7417
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.045886 0.016215 2.830 0.004676 **
vote_count 0.532746 0.012983 41.033 < 2e-16 ***
budget 0.388321 0.010520 36.911 < 2e-16 ***
genre_Family1 0.082813 0.029208 2.835 0.004597 **
holiday_month1 0.123720 0.016236 7.620 3.04e-14 ***
topDirector1 -0.156691 0.028239 -5.549 3.03e-08 ***
genre_Science.Fiction1 -0.168836 0.025145 -6.714 2.11e-11 ***
topActor1 -0.091977 0.017533 -5.246 1.62e-07 ***
genre_Crime1 -0.077725 0.022055 -3.524 0.000429 ***
country_Germany1 -0.112386 0.029560 -3.802 0.000145 ***
popularity 0.040446 0.011808 3.425 0.000619 ***
genre_Animation1 0.141833 0.041666 3.404 0.000669 ***
genre_Western1 -0.206501 0.057265 -3.606 0.000314 ***
genre_Adventure1 0.089030 0.023665 3.762 0.000171 ***
genre_Romance1 0.062104 0.019781 3.140 0.001702 **
genre_Drama1 -0.062777 0.016887 -3.717 0.000204 ***
genre_Action1 -0.058691 0.020400 -2.877 0.004033 **
runtime 0.022660 0.008585 2.639 0.008330 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.5098 on 4783 degrees of freedom
Multiple R-squared: 0.741, Adjusted R-squared: 0.7401
F-statistic: 805.1 on 17 and 4783 DF, p-value: < 2.2e-16dflm2 <- lm(revenue ~ budget + runtime + vote_count + genre_Crime + genre_Drama + genre_Animation + genre_Family + + holiday_month:vote_count + topActor:vote_count + topDirector:vote_count + topDirector:budget + genre_Action:vote_count + genre_Adventure:vote_count + genre_Crime:vote_count + genre_Romance:vote_count + genre_Science.Fiction:vote_count + genre_Western:vote_count + holiday_month + vote_average:vote_count + budget:vote_count + runtime:vote_count + vote_count:popularity , data = tmdb1)
summary(dflm2)
Call:
lm(formula = revenue ~ budget + runtime + vote_count + genre_Crime +
genre_Drama + genre_Animation + genre_Family + +holiday_month:vote_count +
topActor:vote_count + topDirector:vote_count + topDirector:budget +
genre_Action:vote_count + genre_Adventure:vote_count + genre_Crime:vote_count +
genre_Romance:vote_count + genre_Science.Fiction:vote_count +
genre_Western:vote_count + holiday_month + vote_average:vote_count +
budget:vote_count + runtime:vote_count + vote_count:popularity,
data = tmdb1)
Residuals:
Min 1Q Median 3Q Max
-2.9928 -0.1596 -0.0271 0.0987 5.3394
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -0.027209 0.012152 -2.239 0.025201 *
budget 0.227194 0.010568 21.499 < 2e-16 ***
runtime 0.028598 0.007768 3.682 0.000234 ***
vote_count 0.604785 0.019757 30.611 < 2e-16 ***
genre_Crime1 -0.068385 0.019209 -3.560 0.000374 ***
genre_Drama1 -0.063201 0.014650 -4.314 1.63e-05 ***
genre_Animation1 0.198554 0.037992 5.226 1.80e-07 ***
genre_Family1 0.126021 0.025939 4.858 1.22e-06 ***
holiday_month1 0.110184 0.014685 7.503 7.39e-14 ***
vote_count:holiday_month1 0.154833 0.014467 10.702 < 2e-16 ***
vote_count:topActor1 -0.173807 0.015955 -10.893 < 2e-16 ***
vote_count:topDirector1 -0.112115 0.019680 -5.697 1.29e-08 ***
budget:topDirector1 0.100599 0.023552 4.271 1.98e-05 ***
vote_count:genre_Action1 -0.076905 0.016967 -4.533 5.96e-06 ***
vote_count:genre_Adventure1 0.116528 0.018595 6.266 4.02e-10 ***
vote_count:genre_Crime1 -0.126473 0.021737 -5.818 6.33e-09 ***
vote_count:genre_Romance1 0.145046 0.026706 5.431 5.88e-08 ***
vote_count:genre_Science.Fiction1 -0.098211 0.017221 -5.703 1.25e-08 ***
vote_count:genre_Western1 -0.393108 0.047283 -8.314 < 2e-16 ***
vote_count:vote_average -0.085615 0.010699 -8.002 1.52e-15 ***
budget:vote_count 0.095212 0.005404 17.619 < 2e-16 ***
runtime:vote_count 0.045113 0.007598 5.938 3.10e-09 ***
vote_count:popularity -0.008930 0.001951 -4.577 4.85e-06 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.4601 on 4778 degrees of freedom
Multiple R-squared: 0.7893, Adjusted R-squared: 0.7883
F-statistic: 813.5 on 22 and 4778 DF, p-value: < 2.2e-16par(mfrow=c(2,2)) # init 4 charts in 1 panel
plot(dflm2)
par(mfrow = c(1, 2))
hist(dflm1$residuals)
qqnorm(dflm1$residuals)qqline(dflm1$residuals)
library(grid)
library(gridExtra)
Attaching package: ‘gridExtra’
The following object is masked from ‘package:dplyr’:
combineg_budget <- ggplot(data=NULL, aes(x=tmdb1$budget, y=dflm2$residuals)) +
geom_point()
g_votecount <- ggplot(data=NULL, aes(x=tmdb1$vote_count, y=dflm2$residuals)) +
geom_point()
g_popularity <- ggplot(data=NULL, aes(x=tmdb1$popularity, y=dflm2$residuals)) +
geom_point()
grid.arrange(g_budget, g_votecount,g_popularity)
3.1.2. Perfrom stepwise regression to identify the top predictors To explore the data, created several histograms of runtime, vote_count, vote_average, popularity to understand their distribution
options(repr.plot.width=6, repr.plot.height=4)
g1<-ggplot(tmdb1,aes(x=tmdb1$runtime))+geom_histogram(binwidth=5,aes(y=..density..),fill="green4")
g2<-ggplot(tmdb1,aes(x=tmdb1$vote_count))+geom_histogram(binwidth=50,aes(y=..count..),fill="red")
g3<-ggplot(tmdb1,aes(x=tmdb1$popularity))+geom_histogram(binwidth=1,aes(y=..count..),fill="green4")
g4<-ggplot(tmdb1,aes(x=tmdb1$vote_average))+geom_histogram(binwidth=5,aes(y=..count..),fill="red")
grid.arrange(g1,g2,g3,g4,nrow=2,ncol=2)
3.1.3. Split dataset to find test R-squared for linear model
set.seed(1) # for reproducible example
# training set
train <- sample(1:nrow(tmdb1),0.70*nrow(tmdb1)) # random sample of 75% of data
fit <- lm(revenue ~ + budget + runtime + vote_count + genre_Crime + genre_Drama + genre_Animation + genre_Family + + holiday_month:vote_count + topActor:vote_count + topDirector:vote_count + topDirector:budget + genre_Action:vote_count + genre_Adventure:vote_count + genre_Crime:vote_count + genre_Romance:vote_count + genre_Science.Fiction:vote_count + genre_Western:vote_count + holiday_month + vote_average:vote_count + budget:vote_count + runtime:vote_count + vote_count:popularity , data = tmdb1[train,])test <- -train
test.pred <- predict(fit,newdata=tmdb1[test,])
test.y <- tmdb1[test,]$revenue
SS.total <- sum((test.y - mean(test.y))^2)
SS.residual <- sum((test.y - test.pred)^2)
SS.regression <- sum((test.pred - mean(test.y))^2)
SS.total - (SS.regression+SS.residual)[1] 1.270162# [1] 8958890
# NOT the fraction of variability explained by the model
test.rsq <- 1 - SS.residual/SS.total
test.rsq[1] 0.75713093.2. Ridge Regression
x <- model.matrix(revenue ~ budget + runtime + vote_count + genre_Crime + genre_Drama + genre_Animation + genre_Family + + holiday_month:vote_count + topActor:vote_count + topDirector:vote_count + topDirector:budget + genre_Action:vote_count + genre_Adventure:vote_count + genre_Crime:vote_count + genre_Romance:vote_count + genre_Science.Fiction:vote_count + genre_Western:vote_count + holiday_month + vote_average:vote_count + budget:vote_count + runtime:vote_count + vote_count:popularity, tmdb1)
y = tmdb1 %>%
dplyr::select(revenue) %>%
unlist() %>%
as.numeric()3.2.1. Split dataset to train and test
set.seed(1)
train = tmdb1 %>%
sample_frac(0.7)
test = tmdb1 %>%
setdiff(train)
x_train = model.matrix(revenue ~ budget + runtime + vote_count + genre_Crime + genre_Drama + genre_Animation + genre_Family + + holiday_month:vote_count + topActor:vote_count + topDirector:vote_count + topDirector:budget + genre_Action:vote_count + genre_Adventure:vote_count + genre_Crime:vote_count + genre_Romance:vote_count + genre_Science.Fiction:vote_count + genre_Western:vote_count + holiday_month + vote_average:vote_count + budget:vote_count + runtime:vote_count + vote_count:popularity, train)[,-1]
x_test = model.matrix(revenue ~ budget + runtime + vote_count + genre_Crime + genre_Drama + genre_Animation + genre_Family + + holiday_month:vote_count + topActor:vote_count + topDirector:vote_count + topDirector:budget + genre_Action:vote_count + genre_Adventure:vote_count + genre_Crime:vote_count + genre_Romance:vote_count + genre_Science.Fiction:vote_count + genre_Western:vote_count + holiday_month + vote_average:vote_count + budget:vote_count + runtime:vote_count + vote_count:popularity, test)[,-1]
y_train = train %>%
dplyr::select(revenue) %>%
unlist() %>%
as.numeric()
y_test = test %>%
dplyr::select(revenue) %>%
unlist() %>%
as.numeric()3.2.2. Create a set of lambda values and train the model
lambda <- 10^seq(2, -2, length = 100)ridge_mod = glmnet(x_train, y_train, alpha=0, lambda = lambda)
plot(ridge_mod, xvar='lambda', label = TRUE, main = "Ridge Regression")
#Running 10-fold cross validation.
set.seed(0)
cv.ridge.out = cv.glmnet(x_train, y_train, lambda = lambda, alpha = 0, nfolds = 10)
plot(cv.ridge.out, main = "Ridge Regression\n")
bestlambda.ridge = cv.ridge.out$lambda.min
bestlambda.ridge[1] 0.01log(bestlambda.ridge)[1] -4.60517#What is the test MSE associated with this best value of lambda?
ridge.bestlambdatrain = predict(ridge_mod, s = bestlambda.ridge, newx = x_test)
mean((ridge.bestlambdatrain - y_test)^2)[1] 0.2335615y_predicted <- predict(ridge_mod, s = bestlambda.ridge, newx = x_test)
# Sum of Squares Total and Error
sst <- sum((y_test - mean(y_test))^2)
sse <- sum((y_predicted - y_test)^2)
# R squared
rsq_ridge <- 1 - sse / sst
rsq_ridge[1] 0.75832793.2. Lasso Regression
lasso.models = glmnet(x_train, y_train, alpha = 1, lambda = lambda)
plot(lasso.models)
#Running 10-fold cross validation.
set.seed(0)
cv.lasso.out = cv.glmnet(x_train, y_train, lambda = lambda, alpha = 1, nfolds = 10)
plot(cv.lasso.out, main = "Lasso Regression\n")
bestlambda.lasso = cv.lasso.out$lambda.min
bestlambda.lasso[1] 0.01#What is the test MSE associated with this best value of lambda?
lasso.bestlambdatrain = predict(lasso.models, s = bestlambda.lasso, newx = x_test)
mean((lasso.bestlambdatrain - y_test)^2)[1] 0.244208#Fit the lasso regression on the train dataset using the best lambda value
#from cross validation; inspect the coefficient estimates.
lasso.out = glmnet(x_train, y_train, alpha = 1)
predict(lasso.out, type = "coefficients", s = bestlambda.lasso)23 x 1 sparse Matrix of class "dgCMatrix"
1
(Intercept) -0.046415751
budget 0.276864206
runtime 0.008955606
vote_count 0.473686578
genre_Crime1 -0.037614750
genre_Drama1 -0.035471313
genre_Animation1 0.200287305
genre_Family1 0.106017417
holiday_month1 0.094069163
vote_count:holiday_month1 0.147330369
vote_count:topActor1 -0.007233109
vote_count:topDirector1 -0.106619403
budget:topDirector1 0.073059891
vote_count:genre_Action1 .
vote_count:genre_Adventure1 0.077844069
vote_count:genre_Crime1 -0.102668311
vote_count:genre_Romance1 0.195088919
vote_count:genre_Science.Fiction1 -0.113163159
vote_count:genre_Western1 .
vote_count:vote_average -0.069897402
budget:vote_count 0.079538822
runtime:vote_count 0.001773404
vote_count:popularity -0.002170431#Let's also inspect the MSE of our final lasso model on all our data.
lasso.bestlambda = predict(lasso.out, s = bestlambda.lasso, newx = x_test)
mean((lasso.bestlambda - y_test)^2)[1] 0.2442298y_predicted <- predict(lasso.out, s = bestlambda.ridge, newx = x_test)
# Sum of Squares Total and Error
sst <- sum((y_test - mean(y_test))^2)
sse <- sum((y_predicted - y_test)^2)
# R squared
rsq_lasso <- 1 - sse / sst
rsq_lasso[1] 0.74728913.3. Regression Trees
#Create training and testing datasets
library(rpart)
regTree <- rpart(revenue ~ vote_count + budget + holiday_month +
topDirector + topActor + popularity + genre_Family + genre_Crime + genre_Animation +
genre_Adventure + genre_Romance + genre_Drama + genre_Action + genre_Science.Fiction +
runtime, method="anova", data=tmdb1 )plot(regTree, uniform=TRUE,
main="Regression Tree for predicting Revenue")
text(regTree, use.n=TRUE, cex = 0.6)
library(rpart.plot)
rpart.plot(regTree,digits = 4)
p.rpart <- predict(regTree,tmdb1)
mean((p.rpart-tmdb1$revenue)^2)[1] 0.27388# Sum of Squares Total and Error
sst <- sum((tmdb1$revenue - mean(tmdb1$revenue))^2)
sse <- sum((p.rpart - tmdb1$revenue)^2)
# R squared
rsq_tree <- 1 - sse / sst
rsq_tree[1] 0.72606293.4. Random Forest
set.seed(45)
train.index <- sample(row.names(tmdb1), dim(tmdb1)[1]*0.6)
test.index <- setdiff(row.names(tmdb1), train.index)
train <- tmdb1[train.index, ]
test <- tmdb1[test.index, ]set.seed(100)
library(randomForest)
rf <- randomForest(revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector +topActor +popularity + genre_Crime + genre_Animation +
genre_Adventure + genre_Romance + genre_Drama + genre_Action + genre_Science.Fiction +
runtime,data=train,ntree=500)pred_rf <- predict(rf,test)
mean((pred_rf - test$revenue)^2)[1] 0.1856559# Sum of Squares Total and Error
sst <- sum((test$revenue - mean(test$revenue))^2)
sse <- sum((pred_rf - test$revenue)^2)
# R squared
rsq_rf <- 1 - sse / sst
rsq_rf[1] 0.7886412pred_rf1 <- predict(rf,tmdb1)
mean((pred_rf1-tmdb1$revenue)^2)[1] 0.1082728#install reptree packages
have.packages <- installed.packages()
cran.packages <- c('devtools','plotrix','randomForest','tree')
to.install <- setdiff(cran.packages, have.packages[,1])
if(length(to.install)>0) install.packages(to.install)
library(devtools)
if(!('reprtree' %in% installed.packages())){
install_github('araastat/reprtree')
}
for(p in c(cran.packages, 'reprtree')) eval(substitute(library(pkg), list(pkg=p)))library(randomForest)
library(reprtree)
reprtree:::plot.getTree(rf)
set.seed(100)
library(randomForest)
rf_node <- randomForest(revenue ~ vote_count + budget + genre_Family + holiday_month +
topDirector +topActor +popularity + genre_Crime + genre_Animation +
genre_Adventure + genre_Romance + genre_Drama + genre_Action + genre_Science.Fiction +
runtime,data=train, maxnodes=50, ntree=500)reprtree:::plot.getTree(rf_node)
Conclusion
| Predictive Model | R-squared | Comment |
|---|---|---|
| Linear Regression | 0.7571309 | Linear regression has too many non-linearities. The R-squared on complete dataset was 0.789 |
| Ridge Regression | 0.7583279 | Ridge Regression add a penalty on sum of squared beta. This has the effect of “shrinking” large values of beta towards zero. As a result the ridge regression estimates are often more accurate. |
| Lasso Regression | 0.7472891 | The LASSO works in a similar way to ridge regression except that it uses an L1 penalty. LASSO is not quite as computational efficient as ridge regression. |
| Regression Trees | 0.7260629 | A decision tree is built on an entire dataset, using all the features/variables of interest, using all the features/variables of interest |
| Random Froests | 0.7886412 | Accuracy keeps increasing as you increase the number of trees, but becomes constant at certain point. Unlike decision tree, it won’t create highly biased model and reduces the variance. |
Based on our scaled data our top predictors changed in Linear and Random Forest models. In Random Forest we have vote_count, popularity, genre_adventure, top_actor, holiday_month to be top5 predictors.
We conclude that Random Forests is the best model to predict the revenue. The point of RF is to prevent overfitting. It does this by creating random subsets of the features and building smaller (shallow) trees using the subsets and then it combines the subtrees. The downside of RF is it can be slow if you have a single process but it can be parallelized.
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